docs: refresh MCP server guide

Re-work the tool execution flow. Read `orchestrator.rs` to understand
the structure

.github/ISSUE_TEMPLATE/2-bug-report.yml

@@ -20,6 +20,14 @@ body:
     attributes:
       label: What version of Codex is running?
       description: Copy the output of `codex --version`
+    validations:
+      required: true
+  - type: input
+    id: plan
+    attributes:
+      label: What subscription do you have?
+    validations:
+      required: true
   - type: input
     id: model
     attributes:
@@ -32,11 +40,18 @@ body:
       description: |
         For MacOS and Linux: copy the output of `uname -mprs`
         For Windows: copy the output of `"$([Environment]::OSVersion | ForEach-Object VersionString) $(if ([Environment]::Is64BitOperatingSystem) { "x64" } else { "x86" })"` in the PowerShell console
+  - type: textarea
+    id: actual
+    attributes:
+      label: What issue are you seeing?
+      description: Please include the full error messages and prompts with PII redacted. If possible, please provide text instead of a screenshot. 
+    validations:
+      required: true
   - type: textarea
     id: steps
     attributes:
       label: What steps can reproduce the bug?
-      description: Explain the bug and provide a code snippet that can reproduce it.
+      description: Explain the bug and provide a code snippet that can reproduce it. Please include session id, token limit usage, context window usage if applicable.
     validations:
       required: true
   - type: textarea
@@ -44,11 +59,6 @@ body:
     attributes:
       label: What is the expected behavior?
       description: If possible, please provide text instead of a screenshot.
-  - type: textarea
-    id: actual
-    attributes:
-      label: What do you see instead?
-      description: If possible, please provide text instead of a screenshot.
   - type: textarea
     id: notes
     attributes:

.github/ISSUE_TEMPLATE/4-feature-request.yml

@@ -2,7 +2,6 @@ name: 🎁 Feature Request
 description: Propose a new feature for Codex
 labels:
   - enhancement
-  - needs triage
 body:
   - type: markdown
     attributes:
@@ -19,11 +18,6 @@ body:
       label: What feature would you like to see?
     validations:
       required: true
-  - type: textarea
-    id: author
-    attributes:
-      label: Are you interested in implementing this feature?
-      description: Please wait for acknowledgement before implementing or opening a PR.
   - type: textarea
     id: notes
     attributes:

.github/ISSUE_TEMPLATE/5-vs-code-extension.yml

@@ -14,11 +14,21 @@ body:
     id: version
     attributes:
       label: What version of the VS Code extension are you using?
+    validations:
+      required: true
+  - type: input
+    id: plan
+    attributes:
+      label: What subscription do you have?
+    validations:
+      required: true
   - type: input
     id: ide
     attributes:
       label: Which IDE are you using?
       description: Like `VS Code`, `Cursor`, `Windsurf`, etc.
+    validations:
+      required: true
   - type: input
     id: platform
     attributes:
@@ -26,11 +36,18 @@ body:
       description: |
         For MacOS and Linux: copy the output of `uname -mprs`
         For Windows: copy the output of `"$([Environment]::OSVersion | ForEach-Object VersionString) $(if ([Environment]::Is64BitOperatingSystem) { "x64" } else { "x86" })"` in the PowerShell console
+  - type: textarea
+    id: actual
+    attributes:
+      label: What issue are you seeing?
+      description: Please include the full error messages and prompts with PII redacted. If possible, please provide text instead of a screenshot. 
+    validations:
+      required: true
   - type: textarea
     id: steps
     attributes:
       label: What steps can reproduce the bug?
-      description: Explain the bug and provide a code snippet that can reproduce it.
+      description: Explain the bug and provide a code snippet that can reproduce it. Please include session id, token limit usage, context window usage if applicable.
     validations:
       required: true
   - type: textarea
@@ -38,11 +55,6 @@ body:
     attributes:
       label: What is the expected behavior?
       description: If possible, please provide text instead of a screenshot.
-  - type: textarea
-    id: actual
-    attributes:
-      label: What do you see instead?
-      description: If possible, please provide text instead of a screenshot.
   - type: textarea
     id: notes
     attributes:

.github/prompts/issue-deduplicator.txt

@@ -0,0 +1,18 @@
+You are an assistant that triages new GitHub issues by identifying potential duplicates.
+
+You will receive the following JSON files located in the current working directory:
+- `codex-current-issue.json`: JSON object describing the newly created issue (fields: number, title, body).
+- `codex-existing-issues.json`: JSON array of recent issues (each element includes number, title, body, createdAt).
+
+Instructions:
+- Load both files as JSON and review their contents carefully. The codex-existing-issues.json file is large, ensure you explore all of it.
+- Compare the current issue against the existing issues to find up to five that appear to describe the same underlying problem or request.
+- Only consider an issue a potential duplicate if there is a clear overlap in symptoms, feature requests, reproduction steps, or error messages.
+- Prioritize newer issues when similarity is comparable.
+- Ignore pull requests and issues whose similarity is tenuous.
+- When unsure, prefer returning fewer matches.
+
+Output requirements:
+- Respond with a JSON array of issue numbers (integers), ordered from most likely duplicate to least.
+- Include at most five numbers.
+- If you find no plausible duplicates, respond with `[]`.

.github/prompts/issue-labeler.txt

@@ -0,0 +1,26 @@
+You are an assistant that reviews GitHub issues for the repository.
+
+Your job is to choose the most appropriate existing labels for the issue described later in this prompt.
+Follow these rules:
+- Only pick labels out of the list below.
+- Prefer a small set of precise labels over many broad ones.
+- If none of the labels fit, respond with an empty JSON array: []
+- Output must be a JSON array of label names (strings) with no additional commentary.
+
+Labels to apply:
+1. bug — Reproducible defects in Codex products (CLI, VS Code extension, web, auth).
+2. enhancement — Feature requests or usability improvements that ask for new capabilities, better ergonomics, or quality-of-life tweaks.
+3. extension — VS Code (or other IDE) extension-specific issues.
+4. windows-os — Bugs or friction specific to Windows environments (PowerShell behavior, path handling, copy/paste, OS-specific auth or tooling failures).
+5. mcp — Topics involving Model Context Protocol servers/clients.
+6. codex-web — Issues targeting the Codex web UI/Cloud experience.
+8. azure — Problems or requests tied to Azure OpenAI deployments.
+9. documentation — Updates or corrections needed in docs/README/config references (broken links, missing examples, outdated keys, clarification requests).
+10. model-behavior — Undesirable LLM behavior: forgetting goals, refusing work, hallucinating environment details, quota misreports, or other reasoning/performance anomalies.
+
+Issue information is available in environment variables:
+
+ISSUE_NUMBER
+ISSUE_TITLE
+ISSUE_BODY
+REPO_FULL_NAME

.github/workflows/ci.yml

@@ -60,3 +60,6 @@ jobs:
         run: ./scripts/asciicheck.py codex-cli/README.md
       - name: Check codex-cli/README ToC
         run: python3 scripts/readme_toc.py codex-cli/README.md
+
+      - name: Prettier (run `pnpm run format:fix` to fix)
+        run: pnpm run format

.github/workflows/issue-deduplicator.yml

@@ -0,0 +1,140 @@
+name: Issue Deduplicator
+
+on:
+  issues:
+    types:
+      - opened
+      - labeled
+
+jobs:
+  gather-duplicates:
+    name: Identify potential duplicates
+    if: ${{ github.event.action == 'opened' || (github.event.action == 'labeled' && github.event.label.name == 'codex-deduplicate') }}
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    outputs:
+      codex_output: ${{ steps.codex.outputs.final-message }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Prepare Codex inputs
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: |
+          set -eo pipefail
+
+          CURRENT_ISSUE_FILE=codex-current-issue.json
+          EXISTING_ISSUES_FILE=codex-existing-issues.json
+
+          gh issue list --repo "${{ github.repository }}" \
+            --json number,title,body,createdAt \
+            --limit 1000 \
+            --state all \
+            --search "sort:created-desc" \
+            | jq '.' \
+            > "$EXISTING_ISSUES_FILE"
+
+          gh issue view "${{ github.event.issue.number }}" \
+            --repo "${{ github.repository }}" \
+            --json number,title,body \
+            | jq '.' \
+            > "$CURRENT_ISSUE_FILE"
+
+      - id: codex
+        uses: openai/codex-action@main
+        with:
+          openai-api-key: ${{ secrets.CODEX_OPENAI_API_KEY }}
+          allow-users: "*"
+          model: gpt-5
+          prompt: |
+            You are an assistant that triages new GitHub issues by identifying potential duplicates.
+
+            You will receive the following JSON files located in the current working directory:
+            - `codex-current-issue.json`: JSON object describing the newly created issue (fields: number, title, body).
+            - `codex-existing-issues.json`: JSON array of recent issues (each element includes number, title, body, createdAt).
+
+            Instructions:
+            - Compare the current issue against the existing issues to find up to five that appear to describe the same underlying problem or request.
+            - Focus on the underlying intent and context of each issue—such as reported symptoms, feature requests, reproduction steps, or error messages—rather than relying solely on string similarity or synthetic metrics.
+            - After your analysis, validate your results in 1-2 lines explaining your decision to return the selected matches.
+            - When unsure, prefer returning fewer matches.
+            - Include at most five numbers.
+
+          output-schema: |
+            {
+              "type": "object",
+              "properties": {
+                "issues": {
+                  "type": "array",
+                  "items": {
+                    "type": "string"
+                  }
+                },
+                "reason": { "type": "string" }
+              },
+              "required": ["issues", "reason"],
+              "additionalProperties": false
+            }
+
+  comment-on-issue:
+    name: Comment with potential duplicates
+    needs: gather-duplicates
+    if: ${{ needs.gather-duplicates.result != 'skipped' }}
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      issues: write
+    steps:
+      - name: Comment on issue
+        uses: actions/github-script@v7
+        env:
+          CODEX_OUTPUT: ${{ needs.gather-duplicates.outputs.codex_output }}
+        with:
+          github-token: ${{ github.token }}
+          script: |
+            const raw = process.env.CODEX_OUTPUT ?? '';
+            let parsed;
+            try {
+              parsed = JSON.parse(raw);
+            } catch (error) {
+              core.info(`Codex output was not valid JSON. Raw output: ${raw}`);
+              core.info(`Parse error: ${error.message}`);
+              return;
+            }
+
+            const issues = Array.isArray(parsed?.issues) ? parsed.issues : [];
+            const currentIssueNumber = String(context.payload.issue.number);
+
+            console.log(`Current issue number: ${currentIssueNumber}`);
+            console.log(issues);
+
+            const filteredIssues = issues.filter((value) => String(value) !== currentIssueNumber);
+
+            if (filteredIssues.length === 0) {
+              core.info('Codex reported no potential duplicates.');
+              return;
+            }
+
+            const lines = [
+              'Potential duplicates detected. Please review them and close your issue if it is a duplicate.',
+              '',
+              ...filteredIssues.map((value) => `- #${String(value)}`),
+              '',
+              '*Powered by [Codex Action](https://github.com/openai/codex-action)*'];
+
+            await github.rest.issues.createComment({
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              issue_number: context.payload.issue.number,
+              body: lines.join("\n"),
+            });
+
+      - name: Remove codex-deduplicate label
+        if: ${{ always() && github.event.action == 'labeled' && github.event.label.name == 'codex-deduplicate' }}
+        env:
+          GH_TOKEN: ${{ github.token }}
+          GH_REPO: ${{ github.repository }}
+        run: |
+          gh issue edit "${{ github.event.issue.number }}" --remove-label codex-deduplicate || true
+          echo "Attempted to remove label: codex-deduplicate"

.github/workflows/issue-labeler.yml

@@ -0,0 +1,115 @@
+name: Issue Labeler
+
+on:
+  issues:
+    types:
+      - opened
+      - labeled
+
+jobs:
+  gather-labels:
+    name: Generate label suggestions
+    if: ${{ github.event.action == 'opened' || (github.event.action == 'labeled' && github.event.label.name == 'codex-label') }}
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+    outputs:
+      codex_output: ${{ steps.codex.outputs.final-message }}
+    steps:
+      - uses: actions/checkout@v4
+
+      - id: codex
+        uses: openai/codex-action@main
+        with:
+          openai-api-key: ${{ secrets.CODEX_OPENAI_API_KEY }}
+          allow-users: "*"
+          prompt: |
+            You are an assistant that reviews GitHub issues for the repository.
+
+            Your job is to choose the most appropriate existing labels for the issue described later in this prompt.
+            Follow these rules:
+            - Only pick labels out of the list below.
+            - Prefer a small set of precise labels over many broad ones.
+
+            Labels to apply:
+            1. bug — Reproducible defects in Codex products (CLI, VS Code extension, web, auth).
+            2. enhancement — Feature requests or usability improvements that ask for new capabilities, better ergonomics, or quality-of-life tweaks.
+            3. extension — VS Code (or other IDE) extension-specific issues.
+            4. windows-os — Bugs or friction specific to Windows environments (always when PowerShell is mentioned, path handling, copy/paste, OS-specific auth or tooling failures).
+            5. mcp — Topics involving Model Context Protocol servers/clients.
+            6. codex-web — Issues targeting the Codex web UI/Cloud experience.
+            8. azure — Problems or requests tied to Azure OpenAI deployments.
+            9. documentation — Updates or corrections needed in docs/README/config references (broken links, missing examples, outdated keys, clarification requests).
+            10. model-behavior — Undesirable LLM behavior: forgetting goals, refusing work, hallucinating environment details, quota misreports, or other reasoning/performance anomalies.
+
+            Issue number: ${{ github.event.issue.number }}
+
+            Issue title:
+            ${{ github.event.issue.title }}
+
+            Issue body:
+            ${{ github.event.issue.body }}
+
+            Repository full name:
+            ${{ github.repository }}
+
+          output-schema: |
+            {
+              "type": "object",
+              "properties": {
+                "labels": {
+                  "type": "array",
+                  "items": {
+                    "type": "string"
+                  }
+                }
+              },
+              "required": ["labels"],
+              "additionalProperties": false
+            }
+
+  apply-labels:
+    name: Apply labels from Codex output
+    needs: gather-labels
+    if: ${{ needs.gather-labels.result != 'skipped' }}
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      issues: write
+    env:
+      GH_TOKEN: ${{ github.token }}
+      GH_REPO: ${{ github.repository }}
+      ISSUE_NUMBER: ${{ github.event.issue.number }}
+      CODEX_OUTPUT: ${{ needs.gather-labels.outputs.codex_output }}
+    steps:
+      - name: Apply labels
+        run: |
+          json=${CODEX_OUTPUT//$'\r'/}
+          if [ -z "$json" ]; then
+            echo "Codex produced no output. Skipping label application."
+            exit 0
+          fi
+
+          if ! printf '%s' "$json" | jq -e 'type == "object" and (.labels | type == "array")' >/dev/null 2>&1; then
+            echo "Codex output did not include a labels array. Raw output: $json"
+            exit 0
+          fi
+
+          labels=$(printf '%s' "$json" | jq -r '.labels[] | tostring')
+          if [ -z "$labels" ]; then
+            echo "Codex returned an empty array. Nothing to do."
+            exit 0
+          fi
+
+          cmd=(gh issue edit "$ISSUE_NUMBER")
+          while IFS= read -r label; do
+            cmd+=(--add-label "$label")
+          done <<< "$labels"
+
+          "${cmd[@]}" || true
+
+      - name: Remove codex-label trigger
+        if: ${{ always() && github.event.action == 'labeled' && github.event.label.name == 'codex-label' }}
+        run: |
+          gh issue edit "$ISSUE_NUMBER" --remove-label codex-label || true
+          echo "Attempted to remove label: codex-label"

.github/workflows/rust-ci.yml

@@ -148,15 +148,26 @@ jobs:
           targets: ${{ matrix.target }}
           components: clippy
 
-      - uses: actions/cache@v4
+      # Explicit cache restore: split cargo home vs target, so we can
+      # avoid caching the large target dir on the gnu-dev job.
+      - name: Restore cargo home cache
+        id: cache_cargo_home_restore
+        uses: actions/cache/restore@v4
         with:
           path: |
             ~/.cargo/bin/
             ~/.cargo/registry/index/
             ~/.cargo/registry/cache/
             ~/.cargo/git/db/
-            ${{ github.workspace }}/codex-rs/target/
-          key: cargo-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}
+          key: cargo-home-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}
+
+      - name: Restore target cache (except gnu-dev)
+        id: cache_target_restore
+        if: ${{ !(matrix.target == 'x86_64-unknown-linux-gnu' && matrix.profile != 'release') }}
+        uses: actions/cache/restore@v4
+        with:
+          path: ${{ github.workspace }}/codex-rs/target/
+          key: cargo-target-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}
 
       - if: ${{ matrix.target == 'x86_64-unknown-linux-musl' || matrix.target == 'aarch64-unknown-linux-musl'}}
         name: Install musl build tools
@@ -194,6 +205,31 @@ jobs:
         env:
           RUST_BACKTRACE: 1
 
+      # Save caches explicitly; make non-fatal so cache packaging
+      # never fails the overall job. Only save when key wasn't hit.
+      - name: Save cargo home cache
+        if: always() && !cancelled() && steps.cache_cargo_home_restore.outputs.cache-hit != 'true'
+        continue-on-error: true
+        uses: actions/cache/save@v4
+        with:
+          path: |
+            ~/.cargo/bin/
+            ~/.cargo/registry/index/
+            ~/.cargo/registry/cache/
+            ~/.cargo/git/db/
+          key: cargo-home-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}
+
+      - name: Save target cache (except gnu-dev)
+        if: >-
+          always() && !cancelled() &&
+          (steps.cache_target_restore.outputs.cache-hit != 'true') &&
+          !(matrix.target == 'x86_64-unknown-linux-gnu' && matrix.profile != 'release')
+        continue-on-error: true
+        uses: actions/cache/save@v4
+        with:
+          path: ${{ github.workspace }}/codex-rs/target/
+          key: cargo-target-${{ matrix.runner }}-${{ matrix.target }}-${{ matrix.profile }}-${{ hashFiles('**/Cargo.lock') }}
+
       # Fail the job if any of the previous steps failed.
       - name: verify all steps passed
         if: |

.github/workflows/rust-release.yml

@@ -47,7 +47,7 @@ jobs:
 
   build:
     needs: tag-check
-    name: ${{ matrix.runner }} - ${{ matrix.target }}
+    name: Build - ${{ matrix.runner }} - ${{ matrix.target }}
     runs-on: ${{ matrix.runner }}
     timeout-minutes: 30
     defaults:
@@ -94,11 +94,181 @@ jobs:
       - if: ${{ matrix.target == 'x86_64-unknown-linux-musl' || matrix.target == 'aarch64-unknown-linux-musl'}}
         name: Install musl build tools
         run: |
-          sudo apt install -y musl-tools pkg-config
+          sudo apt-get update
+          sudo apt-get install -y musl-tools pkg-config
 
       - name: Cargo build
         run: cargo build --target ${{ matrix.target }} --release --bin codex --bin codex-responses-api-proxy
 
+      - if: ${{ matrix.runner == 'macos-14' }}
+        name: Configure Apple code signing
+        shell: bash
+        env:
+          KEYCHAIN_PASSWORD: actions
+          APPLE_CERTIFICATE: ${{ secrets.APPLE_CERTIFICATE_P12 }}
+          APPLE_CERTIFICATE_PASSWORD: ${{ secrets.APPLE_CERTIFICATE_PASSWORD }}
+        run: |
+          set -euo pipefail
+
+          if [[ -z "${APPLE_CERTIFICATE:-}" ]]; then
+            echo "APPLE_CERTIFICATE is required for macOS signing"
+            exit 1
+          fi
+
+          if [[ -z "${APPLE_CERTIFICATE_PASSWORD:-}" ]]; then
+            echo "APPLE_CERTIFICATE_PASSWORD is required for macOS signing"
+            exit 1
+          fi
+
+          cert_path="${RUNNER_TEMP}/apple_signing_certificate.p12"
+          echo "$APPLE_CERTIFICATE" | base64 -d > "$cert_path"
+
+          keychain_path="${RUNNER_TEMP}/codex-signing.keychain-db"
+          security create-keychain -p "$KEYCHAIN_PASSWORD" "$keychain_path"
+          security set-keychain-settings -lut 21600 "$keychain_path"
+          security unlock-keychain -p "$KEYCHAIN_PASSWORD" "$keychain_path"
+
+          keychain_args=()
+          cleanup_keychain() {
+            if ((${#keychain_args[@]} > 0)); then
+              security list-keychains -s "${keychain_args[@]}" || true
+              security default-keychain -s "${keychain_args[0]}" || true
+            else
+              security list-keychains -s || true
+            fi
+            if [[ -f "$keychain_path" ]]; then
+              security delete-keychain "$keychain_path" || true
+            fi
+          }
+
+          while IFS= read -r keychain; do
+            [[ -n "$keychain" ]] && keychain_args+=("$keychain")
+          done < <(security list-keychains | sed 's/^[[:space:]]*//;s/[[:space:]]*$//;s/"//g')
+
+          if ((${#keychain_args[@]} > 0)); then
+            security list-keychains -s "$keychain_path" "${keychain_args[@]}"
+          else
+            security list-keychains -s "$keychain_path"
+          fi
+
+          security default-keychain -s "$keychain_path"
+          security import "$cert_path" -k "$keychain_path" -P "$APPLE_CERTIFICATE_PASSWORD" -T /usr/bin/codesign -T /usr/bin/security
+          security set-key-partition-list -S apple-tool:,apple: -s -k "$KEYCHAIN_PASSWORD" "$keychain_path" > /dev/null
+
+          codesign_hashes=()
+          while IFS= read -r hash; do
+            [[ -n "$hash" ]] && codesign_hashes+=("$hash")
+          done < <(security find-identity -v -p codesigning "$keychain_path" \
+            | sed -n 's/.*\([0-9A-F]\{40\}\).*/\1/p' \
+            | sort -u)
+
+          if ((${#codesign_hashes[@]} == 0)); then
+            echo "No signing identities found in $keychain_path"
+            cleanup_keychain
+            rm -f "$cert_path"
+            exit 1
+          fi
+
+          if ((${#codesign_hashes[@]} > 1)); then
+            echo "Multiple signing identities found in $keychain_path:"
+            printf '  %s\n' "${codesign_hashes[@]}"
+            cleanup_keychain
+            rm -f "$cert_path"
+            exit 1
+          fi
+
+          APPLE_CODESIGN_IDENTITY="${codesign_hashes[0]}"
+
+          rm -f "$cert_path"
+
+          echo "APPLE_CODESIGN_IDENTITY=$APPLE_CODESIGN_IDENTITY" >> "$GITHUB_ENV"
+          echo "APPLE_CODESIGN_KEYCHAIN=$keychain_path" >> "$GITHUB_ENV"
+          echo "::add-mask::$APPLE_CODESIGN_IDENTITY"
+
+      - if: ${{ matrix.runner == 'macos-14' }}
+        name: Sign macOS binaries
+        shell: bash
+        run: |
+          set -euo pipefail
+
+          if [[ -z "${APPLE_CODESIGN_IDENTITY:-}" ]]; then
+            echo "APPLE_CODESIGN_IDENTITY is required for macOS signing"
+            exit 1
+          fi
+
+          keychain_args=()
+          if [[ -n "${APPLE_CODESIGN_KEYCHAIN:-}" && -f "${APPLE_CODESIGN_KEYCHAIN}" ]]; then
+            keychain_args+=(--keychain "${APPLE_CODESIGN_KEYCHAIN}")
+          fi
+
+          for binary in codex codex-responses-api-proxy; do
+            path="target/${{ matrix.target }}/release/${binary}"
+            codesign --force --options runtime --timestamp --sign "$APPLE_CODESIGN_IDENTITY" "${keychain_args[@]}" "$path"
+          done
+
+      - if: ${{ matrix.runner == 'macos-14' }}
+        name: Notarize macOS binaries
+        shell: bash
+        env:
+          APPLE_NOTARIZATION_KEY_P8: ${{ secrets.APPLE_NOTARIZATION_KEY_P8 }}
+          APPLE_NOTARIZATION_KEY_ID: ${{ secrets.APPLE_NOTARIZATION_KEY_ID }}
+          APPLE_NOTARIZATION_ISSUER_ID: ${{ secrets.APPLE_NOTARIZATION_ISSUER_ID }}
+        run: |
+          set -euo pipefail
+
+          for var in APPLE_NOTARIZATION_KEY_P8 APPLE_NOTARIZATION_KEY_ID APPLE_NOTARIZATION_ISSUER_ID; do
+            if [[ -z "${!var:-}" ]]; then
+              echo "$var is required for notarization"
+              exit 1
+            fi
+          done
+
+          notary_key_path="${RUNNER_TEMP}/notarytool.key.p8"
+          echo "$APPLE_NOTARIZATION_KEY_P8" | base64 -d > "$notary_key_path"
+          cleanup_notary() {
+            rm -f "$notary_key_path"
+          }
+          trap cleanup_notary EXIT
+
+          notarize_binary() {
+            local binary="$1"
+            local source_path="target/${{ matrix.target }}/release/${binary}"
+            local archive_path="${RUNNER_TEMP}/${binary}.zip"
+
+            if [[ ! -f "$source_path" ]]; then
+              echo "Binary $source_path not found"
+              exit 1
+            fi
+
+            rm -f "$archive_path"
+            ditto -c -k --keepParent "$source_path" "$archive_path"
+
+            submission_json=$(xcrun notarytool submit "$archive_path" \
+              --key "$notary_key_path" \
+              --key-id "$APPLE_NOTARIZATION_KEY_ID" \
+              --issuer "$APPLE_NOTARIZATION_ISSUER_ID" \
+              --output-format json \
+              --wait)
+
+            status=$(printf '%s\n' "$submission_json" | jq -r '.status // "Unknown"')
+            submission_id=$(printf '%s\n' "$submission_json" | jq -r '.id // ""')
+
+            if [[ -z "$submission_id" ]]; then
+              echo "Failed to retrieve submission ID for $binary"
+              exit 1
+            fi
+
+            echo "::notice title=Notarization::$binary submission ${submission_id} completed with status ${status}"
+
+            if [[ "$status" != "Accepted" ]]; then
+              echo "Notarization failed for ${binary} (submission ${submission_id}, status ${status})"
+              exit 1
+            fi
+          }
+
+          notarize_binary "codex"
+          notarize_binary "codex-responses-api-proxy"
+
       - name: Stage artifacts
         shell: bash
         run: |
@@ -157,6 +327,29 @@ jobs:
             zstd -T0 -19 --rm "$dest/$base"
           done
 
+      - name: Remove signing keychain
+        if: ${{ always() && matrix.runner == 'macos-14' }}
+        shell: bash
+        env:
+          APPLE_CODESIGN_KEYCHAIN: ${{ env.APPLE_CODESIGN_KEYCHAIN }}
+        run: |
+          set -euo pipefail
+          if [[ -n "${APPLE_CODESIGN_KEYCHAIN:-}" ]]; then
+            keychain_args=()
+            while IFS= read -r keychain; do
+              [[ "$keychain" == "$APPLE_CODESIGN_KEYCHAIN" ]] && continue
+              [[ -n "$keychain" ]] && keychain_args+=("$keychain")
+            done < <(security list-keychains | sed 's/^[[:space:]]*//;s/[[:space:]]*$//;s/"//g')
+            if ((${#keychain_args[@]} > 0)); then
+              security list-keychains -s "${keychain_args[@]}"
+              security default-keychain -s "${keychain_args[0]}"
+            fi
+
+            if [[ -f "$APPLE_CODESIGN_KEYCHAIN" ]]; then
+              security delete-keychain "$APPLE_CODESIGN_KEYCHAIN"
+            fi
+          fi
+
       - uses: actions/upload-artifact@v4
         with:
           name: ${{ matrix.target }}

.gitignore

@@ -30,6 +30,7 @@ result
 # cli tools
 CLAUDE.md
 .claude/
+AGENTS.override.md
 
 # caches
 .cache/

.vscode/settings.json

@@ -3,6 +3,7 @@
     "rust-analyzer.check.command": "clippy",
     "rust-analyzer.check.extraArgs": ["--all-features", "--tests"],
     "rust-analyzer.rustfmt.extraArgs": ["--config", "imports_granularity=Item"],
+    "rust-analyzer.cargo.targetDir": "${workspaceFolder}/codex-rs/target/rust-analyzer",
     "[rust]": {
         "editor.defaultFormatter": "rust-lang.rust-analyzer",
         "editor.formatOnSave": true,

AGENTS.md

@@ -8,11 +8,18 @@ In the codex-rs folder where the rust code lives:
 - Never add or modify any code related to `CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR` or `CODEX_SANDBOX_ENV_VAR`.
   - You operate in a sandbox where `CODEX_SANDBOX_NETWORK_DISABLED=1` will be set whenever you use the `shell` tool. Any existing code that uses `CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR` was authored with this fact in mind. It is often used to early exit out of tests that the author knew you would not be able to run given your sandbox limitations.
   - Similarly, when you spawn a process using Seatbelt (`/usr/bin/sandbox-exec`), `CODEX_SANDBOX=seatbelt` will be set on the child process. Integration tests that want to run Seatbelt themselves cannot be run under Seatbelt, so checks for `CODEX_SANDBOX=seatbelt` are also often used to early exit out of tests, as appropriate.
+- Always collapse if statements per https://rust-lang.github.io/rust-clippy/master/index.html#collapsible_if
+- Always inline format! args when possible per https://rust-lang.github.io/rust-clippy/master/index.html#uninlined_format_args
+- Use method references over closures when possible per https://rust-lang.github.io/rust-clippy/master/index.html#redundant_closure_for_method_calls
+- Do not use unsigned integer even if the number cannot be negative.
+- When writing tests, prefer comparing the equality of entire objects over fields one by one.
+- When making a change that adds or changes an API, ensure that the documentation in the `docs/` folder is up to date if applicable.
 
 Run `just fmt` (in `codex-rs` directory) automatically after making Rust code changes; do not ask for approval to run it. Before finalizing a change to `codex-rs`, run `just fix -p <project>` (in `codex-rs` directory) to fix any linter issues in the code. Prefer scoping with `-p` to avoid slow workspace‑wide Clippy builds; only run `just fix` without `-p` if you changed shared crates. Additionally, run the tests:
+
 1. Run the test for the specific project that was changed. For example, if changes were made in `codex-rs/tui`, run `cargo test -p codex-tui`.
 2. Once those pass, if any changes were made in common, core, or protocol, run the complete test suite with `cargo test --all-features`.
-When running interactively, ask the user before running `just fix` to finalize. `just fmt` does not require approval. project-specific or individual tests can be run without asking the user, but do ask the user before running the complete test suite.
+   When running interactively, ask the user before running `just fix` to finalize. `just fmt` does not require approval. project-specific or individual tests can be run without asking the user, but do ask the user before running the complete test suite.
 
 ## TUI style conventions
 
@@ -28,6 +35,7 @@ See `codex-rs/tui/styles.md`.
     - Desired: vec!["  └ ".into(), "M".red(), " ".dim(), "tui/src/app.rs".dim()]
 
 ### TUI Styling (ratatui)
+
 - Prefer Stylize helpers: use "text".dim(), .bold(), .cyan(), .italic(), .underlined() instead of manual Style where possible.
 - Prefer simple conversions: use "text".into() for spans and vec![…].into() for lines; when inference is ambiguous (e.g., Paragraph::new/Cell::from), use Line::from(spans) or Span::from(text).
 - Computed styles: if the Style is computed at runtime, using `Span::styled` is OK (`Span::from(text).set_style(style)` is also acceptable).
@@ -39,6 +47,7 @@ See `codex-rs/tui/styles.md`.
 - Compactness: prefer the form that stays on one line after rustfmt; if only one of Line::from(vec![…]) or vec![…].into() avoids wrapping, choose that. If both wrap, pick the one with fewer wrapped lines.
 
 ### Text wrapping
+
 - Always use textwrap::wrap to wrap plain strings.
 - If you have a ratatui Line and you want to wrap it, use the helpers in tui/src/wrapping.rs, e.g. word_wrap_lines / word_wrap_line.
 - If you need to indent wrapped lines, use the initial_indent / subsequent_indent options from RtOptions if you can, rather than writing custom logic.
@@ -60,8 +69,34 @@ This repo uses snapshot tests (via `insta`), especially in `codex-rs/tui`, to va
   - `cargo insta accept -p codex-tui`
 
 If you don’t have the tool:
+
 - `cargo install cargo-insta`
 
 ### Test assertions
 
 - Tests should use pretty_assertions::assert_eq for clearer diffs. Import this at the top of the test module if it isn't already.
+
+### Integration tests (core)
+
+- Prefer the utilities in `core_test_support::responses` when writing end-to-end Codex tests.
+
+- All `mount_sse*` helpers return a `ResponseMock`; hold onto it so you can assert against outbound `/responses` POST bodies.
+- Use `ResponseMock::single_request()` when a test should only issue one POST, or `ResponseMock::requests()` to inspect every captured `ResponsesRequest`.
+- `ResponsesRequest` exposes helpers (`body_json`, `input`, `function_call_output`, `custom_tool_call_output`, `call_output`, `header`, `path`, `query_param`) so assertions can target structured payloads instead of manual JSON digging.
+- Build SSE payloads with the provided `ev_*` constructors and the `sse(...)`.
+
+- Typical pattern:
+
+  ```rust
+  let mock = responses::mount_sse_once(&server, responses::sse(vec![
+      responses::ev_response_created("resp-1"),
+      responses::ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
+      responses::ev_completed("resp-1"),
+  ])).await;
+
+  codex.submit(Op::UserTurn { ... }).await?;
+
+  // Assert request body if needed.
+  let request = mock.single_request();
+  // assert using request.function_call_output(call_id) or request.json_body() or other helpers.
+  ```

README.md

@@ -1,5 +1,4 @@
-
-<p align="center"><code>npm i -g @openai/codex</code><br />or <code>brew install codex</code></p>
+<p align="center"><code>npm i -g @openai/codex</code><br />or <code>brew install --cask codex</code></p>
 
 <p align="center"><strong>Codex CLI</strong> is a coding agent from OpenAI that runs locally on your computer.
 </br>
@@ -25,7 +24,7 @@ npm install -g @openai/codex
 Alternatively, if you use Homebrew:
 
 ```shell
-brew install codex
+brew install --cask codex
 ```
 
 Then simply run `codex` to get started:
@@ -62,8 +61,7 @@ You can also use Codex with an API key, but this requires [additional setup](./d
 
 ### Model Context Protocol (MCP)
 
-Codex CLI supports [MCP servers](./docs/advanced.md#model-context-protocol-mcp). Enable by adding an `mcp_servers` section to your `~/.codex/config.toml`.
-
+Codex can access MCP servers. To configure them, refer to the [config docs](./docs/config.md#mcp_servers).
 
 ### Configuration
 
@@ -77,14 +75,18 @@ Codex CLI supports a rich set of configuration options, with preferences stored
   - [CLI usage](./docs/getting-started.md#cli-usage)
   - [Running with a prompt as input](./docs/getting-started.md#running-with-a-prompt-as-input)
   - [Example prompts](./docs/getting-started.md#example-prompts)
+  - [Custom prompts](./docs/prompts.md)
   - [Memory with AGENTS.md](./docs/getting-started.md#memory-with-agentsmd)
   - [Configuration](./docs/config.md)
 - [**Sandbox & approvals**](./docs/sandbox.md)
 - [**Authentication**](./docs/authentication.md)
   - [Auth methods](./docs/authentication.md#forcing-a-specific-auth-method-advanced)
   - [Login on a "Headless" machine](./docs/authentication.md#connecting-on-a-headless-machine)
+- **Automating Codex**
+  - [GitHub Action](https://github.com/openai/codex-action)
+  - [TypeScript SDK](./sdk/typescript/README.md)
+  - [Non-interactive mode (`codex exec`)](./docs/exec.md)
 - [**Advanced**](./docs/advanced.md)
-  - [Non-interactive / CI mode](./docs/advanced.md#non-interactive--ci-mode)
   - [Tracing / verbose logging](./docs/advanced.md#tracing--verbose-logging)
   - [Model Context Protocol (MCP)](./docs/advanced.md#model-context-protocol-mcp)
 - [**Zero data retention (ZDR)**](./docs/zdr.md)

cliff.toml

@@ -4,7 +4,7 @@
 header = """
 # Changelog
 
-You can install any of these versions: `npm install -g codex@version`
+You can install any of these versions: `npm install -g @openai/codex@<version>`
 """
 
 body = """

codex-cli/bin/codex.js

@@ -80,6 +80,32 @@ function getUpdatedPath(newDirs) {
   return updatedPath;
 }
 
+/**
+ * Use heuristics to detect the package manager that was used to install Codex
+ * in order to give the user a hint about how to update it.
+ */
+function detectPackageManager() {
+  const userAgent = process.env.npm_config_user_agent || "";
+  if (/\bbun\//.test(userAgent)) {
+    return "bun";
+  }
+
+  const execPath = process.env.npm_execpath || "";
+  if (execPath.includes("bun")) {
+    return "bun";
+  }
+
+  if (
+    process.env.BUN_INSTALL ||
+    process.env.BUN_INSTALL_GLOBAL_DIR ||
+    process.env.BUN_INSTALL_BIN_DIR
+  ) {
+    return "bun";
+  }
+
+  return userAgent ? "npm" : null;
+}
+
 const additionalDirs = [];
 const pathDir = path.join(archRoot, "path");
 if (existsSync(pathDir)) {
@@ -87,9 +113,16 @@ if (existsSync(pathDir)) {
 }
 const updatedPath = getUpdatedPath(additionalDirs);
 
+const env = { ...process.env, PATH: updatedPath };
+const packageManagerEnvVar =
+  detectPackageManager() === "bun"
+    ? "CODEX_MANAGED_BY_BUN"
+    : "CODEX_MANAGED_BY_NPM";
+env[packageManagerEnvVar] = "1";
+
 const child = spawn(binaryPath, process.argv.slice(2), {
   stdio: "inherit",
-  env: { ...process.env, PATH: updatedPath, CODEX_MANAGED_BY_NPM: "1" },
+  env,
 });
 
 child.on("error", (err) => {

codex-rs/.gitignore

@@ -1,4 +1,5 @@
 /target/
+/target-*/
 
 # Recommended value of CARGO_TARGET_DIR when using Docker as explained in .devcontainer/README.md.
 /target-amd64/

codex-rs/Cargo.toml

@@ -2,10 +2,12 @@
 members = [
     "backend-client",
     "ansi-escape",
+    "async-utils",
     "app-server",
     "app-server-protocol",
     "apply-patch",
     "arg0",
+    "feedback",
     "codex-backend-openapi-models",
     "cloud-tasks",
     "cloud-tasks-client",
@@ -27,11 +29,14 @@ members = [
     "protocol-ts",
     "rmcp-client",
     "responses-api-proxy",
+    "stdio-to-uds",
     "otel",
     "tui",
     "git-apply",
     "utils/json-to-toml",
     "utils/readiness",
+    "utils/pty",
+    "utils/string",
 ]
 resolver = "2"
 
@@ -51,10 +56,12 @@ codex-app-server = { path = "app-server" }
 codex-app-server-protocol = { path = "app-server-protocol" }
 codex-apply-patch = { path = "apply-patch" }
 codex-arg0 = { path = "arg0" }
+codex-async-utils = { path = "async-utils" }
 codex-chatgpt = { path = "chatgpt" }
 codex-common = { path = "common" }
 codex-core = { path = "core" }
 codex-exec = { path = "exec" }
+codex-feedback = { path = "feedback" }
 codex-file-search = { path = "file-search" }
 codex-git-tooling = { path = "git-tooling" }
 codex-linux-sandbox = { path = "linux-sandbox" }
@@ -68,9 +75,12 @@ codex-protocol = { path = "protocol" }
 codex-protocol-ts = { path = "protocol-ts" }
 codex-responses-api-proxy = { path = "responses-api-proxy" }
 codex-rmcp-client = { path = "rmcp-client" }
+codex-stdio-to-uds = { path = "stdio-to-uds" }
 codex-tui = { path = "tui" }
 codex-utils-json-to-toml = { path = "utils/json-to-toml" }
 codex-utils-readiness = { path = "utils/readiness" }
+codex-utils-pty = { path = "utils/pty" }
+codex-utils-string = { path = "utils/string" }
 core_test_support = { path = "core/tests/common" }
 mcp-types = { path = "mcp-types" }
 mcp_test_support = { path = "mcp-server/tests/common" }
@@ -82,9 +92,11 @@ anyhow = "1"
 arboard = "3"
 askama = "0.12"
 assert_cmd = "2"
+assert_matches = "1.5.0"
 async-channel = "2.3.1"
 async-stream = "0.3.6"
 async-trait = "0.1.89"
+axum = { version = "0.8", default-features = false }
 base64 = "0.22.1"
 bytes = "1.10.1"
 chrono = "0.4.42"
@@ -102,7 +114,7 @@ env-flags = "0.1.1"
 env_logger = "0.11.5"
 escargot = "0.5"
 eventsource-stream = "0.2.3"
-futures = "0.3"
+futures = { version = "0.3", default-features = false }
 icu_decimal = "2.0.0"
 icu_locale_core = "2.0.0"
 ignore = "0.4.23"
@@ -110,6 +122,7 @@ image = { version = "^0.25.8", default-features = false }
 indexmap = "2.6.0"
 insta = "1.43.2"
 itertools = "0.14.0"
+keyring = "3.6"
 landlock = "0.4.1"
 lazy_static = "1"
 libc = "0.2.175"
@@ -117,6 +130,7 @@ log = "0.4"
 maplit = "1.0.2"
 mime_guess = "2.0.5"
 multimap = "0.10.0"
+notify = "8.2.0"
 nucleo-matcher = "0.3.1"
 openssl-sys = "*"
 opentelemetry = "0.30.0"
@@ -138,11 +152,14 @@ rand = "0.9"
 ratatui = "0.29.0"
 regex-lite = "0.1.7"
 reqwest = "0.12"
+rmcp = { version = "0.8.0", default-features = false }
 schemars = "0.8.22"
 seccompiler = "0.5.0"
+sentry = "0.34.0"
 serde = "1"
 serde_json = "1"
 serde_with = "3.14"
+serial_test = "3.2.0"
 sha1 = "0.10.6"
 sha2 = "0.10"
 shlex = "1.3.0"
@@ -153,6 +170,7 @@ strum_macros = "0.27.2"
 supports-color = "3.0.2"
 sys-locale = "0.3.2"
 tempfile = "3.23.0"
+test-log = "0.2.18"
 textwrap = "0.16.2"
 thiserror = "2.0.16"
 time = "0.3"
@@ -168,9 +186,11 @@ tracing = "0.1.41"
 tracing-appender = "0.2.3"
 tracing-subscriber = "0.3.20"
 tracing-test = "0.2.5"
-tree-sitter = "0.25.9"
-tree-sitter-bash = "0.25.0"
+tree-sitter = "0.25.10"
+tree-sitter-bash = "0.25"
+tree-sitter-highlight = "0.25.10"
 ts-rs = "11"
+uds_windows = "1.1.0"
 unicode-segmentation = "1.12.0"
 unicode-width = "0.2"
 url = "2"
@@ -237,5 +257,9 @@ strip = "symbols"
 codegen-units = 1
 
 [patch.crates-io]
+# Uncomment to debug local changes.
 # ratatui = { path = "../../ratatui" }
 ratatui = { git = "https://github.com/nornagon/ratatui", branch = "nornagon-v0.29.0-patch" }
+
+# Uncomment to debug local changes.
+# rmcp = { path = "../../rust-sdk/crates/rmcp" }

codex-rs/README.md

@@ -11,7 +11,12 @@ npm i -g @openai/codex
 codex
 ```
 
-You can also install via Homebrew (`brew install codex`) or download a platform-specific release directly from our [GitHub Releases](https://github.com/openai/codex/releases).
+You can also install via Homebrew (`brew install --cask codex`) or download a platform-specific release directly from our [GitHub Releases](https://github.com/openai/codex/releases).
+
+## Documentation quickstart
+
+- First run with Codex? Follow the walkthrough in [`docs/getting-started.md`](../docs/getting-started.md) for prompts, keyboard shortcuts, and session management.
+- Already shipping with Codex and want deeper control? Jump to [`docs/advanced.md`](../docs/advanced.md) and the configuration reference at [`docs/config.md`](../docs/config.md).
 
 ## What's new in the Rust CLI
 
@@ -23,9 +28,15 @@ Codex supports a rich set of configuration options. Note that the Rust CLI uses
 
 ### Model Context Protocol Support
 
-Codex CLI functions as an MCP client that can connect to MCP servers on startup. See the [`mcp_servers`](../docs/config.md#mcp_servers) section in the configuration documentation for details.
+#### MCP client
 
-It is still experimental, but you can also launch Codex as an MCP _server_ by running `codex mcp-server`. Use the [`@modelcontextprotocol/inspector`](https://github.com/modelcontextprotocol/inspector) to try it out:
+Codex CLI functions as an MCP client that allows the Codex CLI and IDE extension to connect to MCP servers on startup. See the [`configuration documentation`](../docs/config.md#mcp_servers) for details.
+
+#### MCP server (experimental)
+
+Codex can be launched as an MCP _server_ by running `codex mcp-server`. This allows _other_ MCP clients to use Codex as a tool for another agent.
+
+Use the [`@modelcontextprotocol/inspector`](https://github.com/modelcontextprotocol/inspector) to try it out:
 
 ```shell
 npx @modelcontextprotocol/inspector codex mcp-server
@@ -41,39 +52,19 @@ You can enable notifications by configuring a script that is run whenever the ag
 
 To run Codex non-interactively, run `codex exec PROMPT` (you can also pass the prompt via `stdin`) and Codex will work on your task until it decides that it is done and exits. Output is printed to the terminal directly. You can set the `RUST_LOG` environment variable to see more about what's going on.
 
-### Use `@` for file search
-
-Typing `@` triggers a fuzzy-filename search over the workspace root. Use up/down to select among the results and Tab or Enter to replace the `@` with the selected path. You can use Esc to cancel the search.
-
-### Esc–Esc to edit a previous message
-
-When the chat composer is empty, press Esc to prime “backtrack” mode. Press Esc again to open a transcript preview highlighting the last user message; press Esc repeatedly to step to older user messages. Press Enter to confirm and Codex will fork the conversation from that point, trim the visible transcript accordingly, and pre‑fill the composer with the selected user message so you can edit and resubmit it.
-
-In the transcript preview, the footer shows an `Esc edit prev` hint while editing is active.
-
-### `--cd`/`-C` flag
-
-Sometimes it is not convenient to `cd` to the directory you want Codex to use as the "working root" before running Codex. Fortunately, `codex` supports a `--cd` option so you can specify whatever folder you want. You can confirm that Codex is honoring `--cd` by double-checking the **workdir** it reports in the TUI at the start of a new session.
-
-### Shell completions
-
-Generate shell completion scripts via:
-
-```shell
-codex completion bash
-codex completion zsh
-codex completion fish
-```
-
 ### Experimenting with the Codex Sandbox
 
 To test to see what happens when a command is run under the sandbox provided by Codex, we provide the following subcommands in Codex CLI:
 
 ```
 # macOS
-codex debug seatbelt [--full-auto] [COMMAND]...
+codex sandbox macos [--full-auto] [COMMAND]...
 
 # Linux
+codex sandbox linux [--full-auto] [COMMAND]...
+
+# Legacy aliases
+codex debug seatbelt [--full-auto] [COMMAND]...
 codex debug landlock [--full-auto] [COMMAND]...
 ```
 

codex-rs/ansi-escape/src/lib.rs

@@ -3,11 +3,30 @@ use ansi_to_tui::IntoText;
 use ratatui::text::Line;
 use ratatui::text::Text;
 
+// Expand tabs in a best-effort way for transcript rendering.
+// Tabs can interact poorly with left-gutter prefixes in our TUI and CLI
+// transcript views (e.g., `nl` separates line numbers from content with a tab).
+// Replacing tabs with spaces avoids odd visual artifacts without changing
+// semantics for our use cases.
+fn expand_tabs(s: &str) -> std::borrow::Cow<'_, str> {
+    if s.contains('\t') {
+        // Keep it simple: replace each tab with 4 spaces.
+        // We do not try to align to tab stops since most usages (like `nl`)
+        // look acceptable with a fixed substitution and this avoids stateful math
+        // across spans.
+        std::borrow::Cow::Owned(s.replace('\t', "    "))
+    } else {
+        std::borrow::Cow::Borrowed(s)
+    }
+}
+
 /// This function should be used when the contents of `s` are expected to match
 /// a single line. If multiple lines are found, a warning is logged and only the
 /// first line is returned.
 pub fn ansi_escape_line(s: &str) -> Line<'static> {
-    let text = ansi_escape(s);
+    // Normalize tabs to spaces to avoid odd gutter collisions in transcript mode.
+    let s = expand_tabs(s);
+    let text = ansi_escape(&s);
     match text.lines.as_slice() {
         [] => "".into(),
         [only] => only.clone(),

codex-rs/app-server-protocol/Cargo.toml

@@ -11,8 +11,11 @@ path = "src/lib.rs"
 workspace = true
 
 [dependencies]
+anyhow = { workspace = true }
+clap = { workspace = true, features = ["derive"] }
 codex-protocol = { workspace = true }
 paste = { workspace = true }
+schemars = { workspace = true }
 serde = { workspace = true, features = ["derive"] }
 serde_json = { workspace = true }
 strum_macros = { workspace = true }

codex-rs/app-server-protocol/src/bin/export.rs

@@ -0,0 +1,22 @@
+use anyhow::Result;
+use clap::Parser;
+use std::path::PathBuf;
+
+#[derive(Parser, Debug)]
+#[command(
+    about = "Generate TypeScript bindings and JSON Schemas for the Codex app-server protocol"
+)]
+struct Args {
+    /// Output directory where generated files will be written
+    #[arg(short = 'o', long = "out", value_name = "DIR")]
+    out_dir: PathBuf,
+
+    /// Optional Prettier executable path to format generated TypeScript files
+    #[arg(short = 'p', long = "prettier", value_name = "PRETTIER_BIN")]
+    prettier: Option<PathBuf>,
+}
+
+fn main() -> Result<()> {
+    let args = Args::parse();
+    codex_app_server_protocol::generate_types(&args.out_dir, args.prettier.as_deref())
+}

codex-rs/app-server-protocol/src/export.rs

@@ -0,0 +1,404 @@
+use crate::ClientNotification;
+use crate::ClientRequest;
+use crate::ServerNotification;
+use crate::ServerRequest;
+use crate::export_client_response_schemas;
+use crate::export_client_responses;
+use crate::export_server_response_schemas;
+use crate::export_server_responses;
+use anyhow::Context;
+use anyhow::Result;
+use anyhow::anyhow;
+use schemars::JsonSchema;
+use schemars::schema::RootSchema;
+use schemars::schema_for;
+use serde::Serialize;
+use serde_json::Map;
+use serde_json::Value;
+use std::collections::BTreeMap;
+use std::ffi::OsStr;
+use std::fs;
+use std::io::Read;
+use std::io::Write;
+use std::path::Path;
+use std::path::PathBuf;
+use std::process::Command;
+use ts_rs::TS;
+
+const HEADER: &str = "// GENERATED CODE! DO NOT MODIFY BY HAND!\n\n";
+
+macro_rules! for_each_schema_type {
+    ($macro:ident) => {
+        $macro!(crate::RequestId);
+        $macro!(crate::JSONRPCMessage);
+        $macro!(crate::JSONRPCRequest);
+        $macro!(crate::JSONRPCNotification);
+        $macro!(crate::JSONRPCResponse);
+        $macro!(crate::JSONRPCError);
+        $macro!(crate::JSONRPCErrorError);
+        $macro!(crate::AddConversationListenerParams);
+        $macro!(crate::AddConversationSubscriptionResponse);
+        $macro!(crate::ApplyPatchApprovalParams);
+        $macro!(crate::ApplyPatchApprovalResponse);
+        $macro!(crate::ArchiveConversationParams);
+        $macro!(crate::ArchiveConversationResponse);
+        $macro!(crate::AuthMode);
+        $macro!(crate::AuthStatusChangeNotification);
+        $macro!(crate::CancelLoginChatGptParams);
+        $macro!(crate::CancelLoginChatGptResponse);
+        $macro!(crate::ClientInfo);
+        $macro!(crate::ClientNotification);
+        $macro!(crate::ClientRequest);
+        $macro!(crate::ConversationSummary);
+        $macro!(crate::ExecCommandApprovalParams);
+        $macro!(crate::ExecCommandApprovalResponse);
+        $macro!(crate::ExecOneOffCommandParams);
+        $macro!(crate::ExecOneOffCommandResponse);
+        $macro!(crate::FuzzyFileSearchParams);
+        $macro!(crate::FuzzyFileSearchResponse);
+        $macro!(crate::FuzzyFileSearchResult);
+        $macro!(crate::GetAuthStatusParams);
+        $macro!(crate::GetAuthStatusResponse);
+        $macro!(crate::GetUserAgentResponse);
+        $macro!(crate::GetUserSavedConfigResponse);
+        $macro!(crate::GitDiffToRemoteParams);
+        $macro!(crate::GitDiffToRemoteResponse);
+        $macro!(crate::GitSha);
+        $macro!(crate::InitializeParams);
+        $macro!(crate::InitializeResponse);
+        $macro!(crate::InputItem);
+        $macro!(crate::InterruptConversationParams);
+        $macro!(crate::InterruptConversationResponse);
+        $macro!(crate::ListConversationsParams);
+        $macro!(crate::ListConversationsResponse);
+        $macro!(crate::LoginApiKeyParams);
+        $macro!(crate::LoginApiKeyResponse);
+        $macro!(crate::LoginChatGptCompleteNotification);
+        $macro!(crate::LoginChatGptResponse);
+        $macro!(crate::LogoutChatGptParams);
+        $macro!(crate::LogoutChatGptResponse);
+        $macro!(crate::NewConversationParams);
+        $macro!(crate::NewConversationResponse);
+        $macro!(crate::Profile);
+        $macro!(crate::RemoveConversationListenerParams);
+        $macro!(crate::RemoveConversationSubscriptionResponse);
+        $macro!(crate::ResumeConversationParams);
+        $macro!(crate::ResumeConversationResponse);
+        $macro!(crate::SandboxSettings);
+        $macro!(crate::SendUserMessageParams);
+        $macro!(crate::SendUserMessageResponse);
+        $macro!(crate::SendUserTurnParams);
+        $macro!(crate::SendUserTurnResponse);
+        $macro!(crate::ServerNotification);
+        $macro!(crate::ServerRequest);
+        $macro!(crate::SessionConfiguredNotification);
+        $macro!(crate::SetDefaultModelParams);
+        $macro!(crate::SetDefaultModelResponse);
+        $macro!(crate::Tools);
+        $macro!(crate::UserInfoResponse);
+        $macro!(crate::UserSavedConfig);
+        $macro!(codex_protocol::protocol::EventMsg);
+        $macro!(codex_protocol::protocol::FileChange);
+        $macro!(codex_protocol::parse_command::ParsedCommand);
+        $macro!(codex_protocol::protocol::SandboxPolicy);
+    };
+}
+
+pub fn generate_types(out_dir: &Path, prettier: Option<&Path>) -> Result<()> {
+    generate_ts(out_dir, prettier)?;
+    generate_json(out_dir)?;
+    Ok(())
+}
+
+pub fn generate_ts(out_dir: &Path, prettier: Option<&Path>) -> Result<()> {
+    ensure_dir(out_dir)?;
+
+    ClientRequest::export_all_to(out_dir)?;
+    export_client_responses(out_dir)?;
+    ClientNotification::export_all_to(out_dir)?;
+
+    ServerRequest::export_all_to(out_dir)?;
+    export_server_responses(out_dir)?;
+    ServerNotification::export_all_to(out_dir)?;
+
+    generate_index_ts(out_dir)?;
+
+    let ts_files = ts_files_in(out_dir)?;
+    for file in &ts_files {
+        prepend_header_if_missing(file)?;
+    }
+
+    if let Some(prettier_bin) = prettier
+        && !ts_files.is_empty()
+    {
+        let status = Command::new(prettier_bin)
+            .arg("--write")
+            .args(ts_files.iter().map(|p| p.as_os_str()))
+            .status()
+            .with_context(|| format!("Failed to invoke Prettier at {}", prettier_bin.display()))?;
+        if !status.success() {
+            return Err(anyhow!("Prettier failed with status {status}"));
+        }
+    }
+
+    Ok(())
+}
+
+pub fn generate_json(out_dir: &Path) -> Result<()> {
+    ensure_dir(out_dir)?;
+    let mut bundle: BTreeMap<String, RootSchema> = BTreeMap::new();
+
+    macro_rules! add_schema {
+        ($ty:path) => {{
+            let name = type_basename(stringify!($ty));
+            let schema = write_json_schema_with_return::<$ty>(out_dir, &name)?;
+            bundle.insert(name, schema);
+        }};
+    }
+
+    for_each_schema_type!(add_schema);
+
+    export_client_response_schemas(out_dir)?;
+    export_server_response_schemas(out_dir)?;
+
+    let mut definitions = Map::new();
+
+    const SPECIAL_DEFINITIONS: &[&str] = &[
+        "ClientNotification",
+        "ClientRequest",
+        "EventMsg",
+        "FileChange",
+        "InputItem",
+        "ParsedCommand",
+        "SandboxPolicy",
+        "ServerNotification",
+        "ServerRequest",
+    ];
+
+    for (name, schema) in bundle {
+        let mut schema_value = serde_json::to_value(schema)?;
+        if let Value::Object(ref mut obj) = schema_value {
+            if let Some(defs) = obj.remove("definitions")
+                && let Value::Object(defs_obj) = defs
+            {
+                for (def_name, def_schema) in defs_obj {
+                    if !SPECIAL_DEFINITIONS.contains(&def_name.as_str()) {
+                        definitions.insert(def_name, def_schema);
+                    }
+                }
+            }
+
+            if let Some(Value::Array(one_of)) = obj.get_mut("oneOf") {
+                for variant in one_of.iter_mut() {
+                    if let Some(variant_name) = variant_definition_name(&name, variant)
+                        && let Value::Object(variant_obj) = variant
+                    {
+                        variant_obj.insert("title".into(), Value::String(variant_name));
+                    }
+                }
+            }
+        }
+        definitions.insert(name, schema_value);
+    }
+
+    let mut root = Map::new();
+    root.insert(
+        "$schema".to_string(),
+        Value::String("http://json-schema.org/draft-07/schema#".into()),
+    );
+    root.insert(
+        "title".to_string(),
+        Value::String("CodexAppServerProtocol".into()),
+    );
+    root.insert("type".to_string(), Value::String("object".into()));
+    root.insert("definitions".to_string(), Value::Object(definitions));
+
+    write_pretty_json(
+        out_dir.join("codex_app_server_protocol.schemas.json"),
+        &Value::Object(root),
+    )?;
+
+    Ok(())
+}
+
+fn write_json_schema_with_return<T>(out_dir: &Path, name: &str) -> Result<RootSchema>
+where
+    T: JsonSchema,
+{
+    let file_stem = name.trim();
+    let schema = schema_for!(T);
+    write_pretty_json(out_dir.join(format!("{file_stem}.json")), &schema)
+        .with_context(|| format!("Failed to write JSON schema for {file_stem}"))?;
+    Ok(schema)
+}
+
+pub(crate) fn write_json_schema<T>(out_dir: &Path, name: &str) -> Result<()>
+where
+    T: JsonSchema,
+{
+    write_json_schema_with_return::<T>(out_dir, name).map(|_| ())
+}
+
+fn write_pretty_json(path: PathBuf, value: &impl Serialize) -> Result<()> {
+    let json = serde_json::to_vec_pretty(value)
+        .with_context(|| format!("Failed to serialize JSON schema to {}", path.display()))?;
+    fs::write(&path, json).with_context(|| format!("Failed to write {}", path.display()))?;
+    Ok(())
+}
+fn type_basename(type_path: &str) -> String {
+    type_path
+        .rsplit_once("::")
+        .map(|(_, name)| name)
+        .unwrap_or(type_path)
+        .trim()
+        .to_string()
+}
+
+fn variant_definition_name(base: &str, variant: &Value) -> Option<String> {
+    if let Some(props) = variant.get("properties").and_then(Value::as_object) {
+        if let Some(method_literal) = literal_from_property(props, "method") {
+            let pascal = to_pascal_case(method_literal);
+            return Some(match base {
+                "ClientRequest" | "ServerRequest" => format!("{pascal}Request"),
+                "ClientNotification" | "ServerNotification" => format!("{pascal}Notification"),
+                _ => format!("{pascal}{base}"),
+            });
+        }
+
+        if let Some(type_literal) = literal_from_property(props, "type") {
+            let pascal = to_pascal_case(type_literal);
+            return Some(match base {
+                "EventMsg" => format!("{pascal}EventMsg"),
+                _ => format!("{pascal}{base}"),
+            });
+        }
+
+        if let Some(mode_literal) = literal_from_property(props, "mode") {
+            let pascal = to_pascal_case(mode_literal);
+            return Some(match base {
+                "SandboxPolicy" => format!("{pascal}SandboxPolicy"),
+                _ => format!("{pascal}{base}"),
+            });
+        }
+
+        if props.len() == 1
+            && let Some(key) = props.keys().next()
+        {
+            let pascal = to_pascal_case(key);
+            return Some(format!("{pascal}{base}"));
+        }
+    }
+
+    if let Some(required) = variant.get("required").and_then(Value::as_array)
+        && required.len() == 1
+        && let Some(key) = required[0].as_str()
+    {
+        let pascal = to_pascal_case(key);
+        return Some(format!("{pascal}{base}"));
+    }
+
+    None
+}
+
+fn literal_from_property<'a>(props: &'a Map<String, Value>, key: &str) -> Option<&'a str> {
+    props
+        .get(key)
+        .and_then(|value| value.get("enum"))
+        .and_then(Value::as_array)
+        .and_then(|arr| arr.first())
+        .and_then(Value::as_str)
+}
+
+fn to_pascal_case(input: &str) -> String {
+    let mut result = String::new();
+    let mut capitalize_next = true;
+
+    for c in input.chars() {
+        if c == '_' || c == '-' {
+            capitalize_next = true;
+            continue;
+        }
+
+        if capitalize_next {
+            result.extend(c.to_uppercase());
+            capitalize_next = false;
+        } else {
+            result.push(c);
+        }
+    }
+
+    result
+}
+
+fn ensure_dir(dir: &Path) -> Result<()> {
+    fs::create_dir_all(dir)
+        .with_context(|| format!("Failed to create output directory {}", dir.display()))
+}
+
+fn prepend_header_if_missing(path: &Path) -> Result<()> {
+    let mut content = String::new();
+    {
+        let mut f = fs::File::open(path)
+            .with_context(|| format!("Failed to open {} for reading", path.display()))?;
+        f.read_to_string(&mut content)
+            .with_context(|| format!("Failed to read {}", path.display()))?;
+    }
+
+    if content.starts_with(HEADER) {
+        return Ok(());
+    }
+
+    let mut f = fs::File::create(path)
+        .with_context(|| format!("Failed to open {} for writing", path.display()))?;
+    f.write_all(HEADER.as_bytes())
+        .with_context(|| format!("Failed to write header to {}", path.display()))?;
+    f.write_all(content.as_bytes())
+        .with_context(|| format!("Failed to write content to {}", path.display()))?;
+    Ok(())
+}
+
+fn ts_files_in(dir: &Path) -> Result<Vec<PathBuf>> {
+    let mut files = Vec::new();
+    for entry in
+        fs::read_dir(dir).with_context(|| format!("Failed to read dir {}", dir.display()))?
+    {
+        let entry = entry?;
+        let path = entry.path();
+        if path.is_file() && path.extension() == Some(OsStr::new("ts")) {
+            files.push(path);
+        }
+    }
+    files.sort();
+    Ok(files)
+}
+
+fn generate_index_ts(out_dir: &Path) -> Result<PathBuf> {
+    let mut entries: Vec<String> = Vec::new();
+    let mut stems: Vec<String> = ts_files_in(out_dir)?
+        .into_iter()
+        .filter_map(|p| {
+            let stem = p.file_stem()?.to_string_lossy().into_owned();
+            if stem == "index" { None } else { Some(stem) }
+        })
+        .collect();
+    stems.sort();
+    stems.dedup();
+
+    for name in stems {
+        entries.push(format!("export type {{ {name} }} from \"./{name}\";\n"));
+    }
+
+    let mut content =
+        String::with_capacity(HEADER.len() + entries.iter().map(String::len).sum::<usize>());
+    content.push_str(HEADER);
+    for line in &entries {
+        content.push_str(line);
+    }
+
+    let index_path = out_dir.join("index.ts");
+    let mut f = fs::File::create(&index_path)
+        .with_context(|| format!("Failed to create {}", index_path.display()))?;
+    f.write_all(content.as_bytes())
+        .with_context(|| format!("Failed to write {}", index_path.display()))?;
+    Ok(index_path)
+}

codex-rs/app-server-protocol/src/jsonrpc_lite.rs

@@ -1,13 +1,14 @@
 //! We do not do true JSON-RPC 2.0, as we neither send nor expect the
 //! "jsonrpc": "2.0" field.
 
+use schemars::JsonSchema;
 use serde::Deserialize;
 use serde::Serialize;
 use ts_rs::TS;
 
 pub const JSONRPC_VERSION: &str = "2.0";
 
-#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, Hash, Eq, TS)]
+#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, Hash, Eq, JsonSchema, TS)]
 #[serde(untagged)]
 pub enum RequestId {
     String(String),
@@ -18,7 +19,7 @@ pub enum RequestId {
 pub type Result = serde_json::Value;
 
 /// Refers to any valid JSON-RPC object that can be decoded off the wire, or encoded to be sent.
-#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, TS)]
+#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, JsonSchema, TS)]
 #[serde(untagged)]
 pub enum JSONRPCMessage {
     Request(JSONRPCRequest),
@@ -28,7 +29,7 @@ pub enum JSONRPCMessage {
 }
 
 /// A request that expects a response.
-#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, TS)]
+#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, JsonSchema, TS)]
 pub struct JSONRPCRequest {
     pub id: RequestId,
     pub method: String,
@@ -37,7 +38,7 @@ pub struct JSONRPCRequest {
 }
 
 /// A notification which does not expect a response.
-#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, TS)]
+#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, JsonSchema, TS)]
 pub struct JSONRPCNotification {
     pub method: String,
     #[serde(default, skip_serializing_if = "Option::is_none")]
@@ -45,20 +46,20 @@ pub struct JSONRPCNotification {
 }
 
 /// A successful (non-error) response to a request.
-#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, TS)]
+#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, JsonSchema, TS)]
 pub struct JSONRPCResponse {
     pub id: RequestId,
     pub result: Result,
 }
 
 /// A response to a request that indicates an error occurred.
-#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, TS)]
+#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, JsonSchema, TS)]
 pub struct JSONRPCError {
     pub error: JSONRPCErrorError,
     pub id: RequestId,
 }
 
-#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, TS)]
+#[derive(Debug, Clone, PartialEq, Deserialize, Serialize, JsonSchema, TS)]
 pub struct JSONRPCErrorError {
     pub code: i64,
     #[serde(default, skip_serializing_if = "Option::is_none")]

codex-rs/app-server-protocol/src/lib.rs

@@ -1,5 +1,9 @@
+mod export;
 mod jsonrpc_lite;
 mod protocol;
 
+pub use export::generate_json;
+pub use export::generate_ts;
+pub use export::generate_types;
 pub use jsonrpc_lite::*;
 pub use protocol::*;

codex-rs/app-server-protocol/src/protocol.rs

@@ -5,10 +5,12 @@ use crate::JSONRPCNotification;
 use crate::JSONRPCRequest;
 use crate::RequestId;
 use codex_protocol::ConversationId;
+use codex_protocol::config_types::ForcedLoginMethod;
 use codex_protocol::config_types::ReasoningEffort;
 use codex_protocol::config_types::ReasoningSummary;
 use codex_protocol::config_types::SandboxMode;
 use codex_protocol::config_types::Verbosity;
+use codex_protocol::parse_command::ParsedCommand;
 use codex_protocol::protocol::AskForApproval;
 use codex_protocol::protocol::EventMsg;
 use codex_protocol::protocol::FileChange;
@@ -16,13 +18,14 @@ use codex_protocol::protocol::ReviewDecision;
 use codex_protocol::protocol::SandboxPolicy;
 use codex_protocol::protocol::TurnAbortReason;
 use paste::paste;
+use schemars::JsonSchema;
 use serde::Deserialize;
 use serde::Serialize;
 use strum_macros::Display;
 use ts_rs::TS;
 use uuid::Uuid;
 
-#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Clone, Debug, PartialEq, JsonSchema, TS)]
 #[ts(type = "string")]
 pub struct GitSha(pub String);
 
@@ -32,7 +35,7 @@ impl GitSha {
     }
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, Display, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, Display, JsonSchema, TS)]
 #[serde(rename_all = "lowercase")]
 pub enum AuthMode {
     ApiKey,
@@ -54,7 +57,7 @@ macro_rules! client_request_definitions {
         ),* $(,)?
     ) => {
         /// Request from the client to the server.
-        #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+        #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
         #[serde(tag = "method", rename_all = "camelCase")]
         pub enum ClientRequest {
             $(
@@ -76,6 +79,15 @@ macro_rules! client_request_definitions {
             )*
             Ok(())
         }
+
+        pub fn export_client_response_schemas(
+            out_dir: &::std::path::Path,
+        ) -> ::anyhow::Result<()> {
+            $(
+                crate::export::write_json_schema::<$response>(out_dir, stringify!($response))?;
+            )*
+            Ok(())
+        }
     };
 }
 
@@ -173,13 +185,13 @@ client_request_definitions! {
     },
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct InitializeParams {
     pub client_info: ClientInfo,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ClientInfo {
     pub name: String,
@@ -188,13 +200,13 @@ pub struct ClientInfo {
     pub version: String,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct InitializeResponse {
     pub user_agent: String,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct NewConversationParams {
     /// Optional override for the model name (e.g. "o3", "o4-mini").
@@ -237,7 +249,7 @@ pub struct NewConversationParams {
     pub include_apply_patch_tool: Option<bool>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct NewConversationResponse {
     pub conversation_id: ConversationId,
@@ -248,7 +260,7 @@ pub struct NewConversationResponse {
     pub rollout_path: PathBuf,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ResumeConversationResponse {
     pub conversation_id: ConversationId,
@@ -257,7 +269,7 @@ pub struct ResumeConversationResponse {
     pub initial_messages: Option<Vec<EventMsg>>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ListConversationsParams {
     /// Optional page size; defaults to a reasonable server-side value.
@@ -268,7 +280,7 @@ pub struct ListConversationsParams {
     pub cursor: Option<String>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ConversationSummary {
     pub conversation_id: ConversationId,
@@ -279,7 +291,7 @@ pub struct ConversationSummary {
     pub timestamp: Option<String>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ListConversationsResponse {
     pub items: Vec<ConversationSummary>,
@@ -289,7 +301,7 @@ pub struct ListConversationsResponse {
     pub next_cursor: Option<String>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ResumeConversationParams {
     /// Absolute path to the rollout JSONL file.
@@ -299,78 +311,81 @@ pub struct ResumeConversationParams {
     pub overrides: Option<NewConversationParams>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct AddConversationSubscriptionResponse {
+    #[schemars(with = "String")]
     pub subscription_id: Uuid,
 }
 
 /// The [`ConversationId`] must match the `rollout_path`.
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ArchiveConversationParams {
     pub conversation_id: ConversationId,
     pub rollout_path: PathBuf,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ArchiveConversationResponse {}
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct RemoveConversationSubscriptionResponse {}
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct LoginApiKeyParams {
     pub api_key: String,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct LoginApiKeyResponse {}
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct LoginChatGptResponse {
+    #[schemars(with = "String")]
     pub login_id: Uuid,
     /// URL the client should open in a browser to initiate the OAuth flow.
     pub auth_url: String,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct GitDiffToRemoteResponse {
     pub sha: GitSha,
     pub diff: String,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct CancelLoginChatGptParams {
+    #[schemars(with = "String")]
     pub login_id: Uuid,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct GitDiffToRemoteParams {
     pub cwd: PathBuf,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct CancelLoginChatGptResponse {}
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct LogoutChatGptParams {}
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct LogoutChatGptResponse {}
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct GetAuthStatusParams {
     /// If true, include the current auth token (if available) in the response.
@@ -381,7 +396,7 @@ pub struct GetAuthStatusParams {
     pub refresh_token: Option<bool>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ExecOneOffCommandParams {
     /// Command argv to execute.
@@ -397,7 +412,7 @@ pub struct ExecOneOffCommandParams {
     pub sandbox_policy: Option<SandboxPolicy>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ExecOneOffCommandResponse {
     pub exit_code: i32,
@@ -405,7 +420,7 @@ pub struct ExecOneOffCommandResponse {
     pub stderr: String,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct GetAuthStatusResponse {
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -420,13 +435,13 @@ pub struct GetAuthStatusResponse {
     pub requires_openai_auth: Option<bool>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct GetUserAgentResponse {
     pub user_agent: String,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct UserInfoResponse {
     /// Note: `alleged_user_email` is not currently verified. We read it from
@@ -436,13 +451,13 @@ pub struct UserInfoResponse {
     pub alleged_user_email: Option<String>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct GetUserSavedConfigResponse {
     pub config: UserSavedConfig,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct SetDefaultModelParams {
     /// If set to None, this means `model` should be cleared in config.toml.
@@ -454,14 +469,14 @@ pub struct SetDefaultModelParams {
     pub reasoning_effort: Option<ReasoningEffort>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct SetDefaultModelResponse {}
 
 /// UserSavedConfig contains a subset of the config. It is meant to expose mcp
 /// client-configurable settings that can be specified in the NewConversation
 /// and SendUserTurn requests.
-#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, TS)]
+#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct UserSavedConfig {
     /// Approvals
@@ -472,6 +487,11 @@ pub struct UserSavedConfig {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub sandbox_settings: Option<SandboxSettings>,
 
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub forced_chatgpt_workspace_id: Option<String>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub forced_login_method: Option<ForcedLoginMethod>,
+
     /// Model-specific configuration
     #[serde(skip_serializing_if = "Option::is_none")]
     pub model: Option<String>,
@@ -494,7 +514,7 @@ pub struct UserSavedConfig {
 }
 
 /// MCP representation of a [`codex_core::config_profile::ConfigProfile`].
-#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, TS)]
+#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct Profile {
     pub model: Option<String>,
@@ -508,7 +528,7 @@ pub struct Profile {
     pub chatgpt_base_url: Option<String>,
 }
 /// MCP representation of a [`codex_core::config::ToolsToml`].
-#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, TS)]
+#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct Tools {
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -518,7 +538,7 @@ pub struct Tools {
 }
 
 /// MCP representation of a [`codex_core::config_types::SandboxWorkspaceWrite`].
-#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, TS)]
+#[derive(Deserialize, Debug, Clone, PartialEq, Serialize, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct SandboxSettings {
     #[serde(default)]
@@ -531,14 +551,14 @@ pub struct SandboxSettings {
     pub exclude_slash_tmp: Option<bool>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct SendUserMessageParams {
     pub conversation_id: ConversationId,
     pub items: Vec<InputItem>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct SendUserTurnParams {
     pub conversation_id: ConversationId,
@@ -552,39 +572,40 @@ pub struct SendUserTurnParams {
     pub summary: ReasoningSummary,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct SendUserTurnResponse {}
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct InterruptConversationParams {
     pub conversation_id: ConversationId,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct InterruptConversationResponse {
     pub abort_reason: TurnAbortReason,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct SendUserMessageResponse {}
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct AddConversationListenerParams {
     pub conversation_id: ConversationId,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct RemoveConversationListenerParams {
+    #[schemars(with = "String")]
     pub subscription_id: Uuid,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[serde(tag = "type", content = "data")]
 pub enum InputItem {
@@ -616,7 +637,7 @@ macro_rules! server_request_definitions {
     ) => {
         paste! {
             /// Request initiated from the server and sent to the client.
-            #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+            #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
             #[serde(tag = "method", rename_all = "camelCase")]
             pub enum ServerRequest {
                 $(
@@ -629,7 +650,7 @@ macro_rules! server_request_definitions {
                 )*
             }
 
-            #[derive(Debug, Clone, PartialEq)]
+            #[derive(Debug, Clone, PartialEq, JsonSchema)]
             pub enum ServerRequestPayload {
                 $( $variant([<$variant Params>]), )*
             }
@@ -651,6 +672,15 @@ macro_rules! server_request_definitions {
             }
             Ok(())
         }
+
+        pub fn export_server_response_schemas(
+            out_dir: &::std::path::Path,
+        ) -> ::anyhow::Result<()> {
+            paste! {
+                $(crate::export::write_json_schema::<[<$variant Response>]>(out_dir, stringify!([<$variant Response>]))?;)*
+            }
+            Ok(())
+        }
     };
 }
 
@@ -669,7 +699,7 @@ server_request_definitions! {
     ExecCommandApproval,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ApplyPatchApprovalParams {
     pub conversation_id: ConversationId,
@@ -686,7 +716,7 @@ pub struct ApplyPatchApprovalParams {
     pub grant_root: Option<PathBuf>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct ExecCommandApprovalParams {
     pub conversation_id: ConversationId,
@@ -697,19 +727,20 @@ pub struct ExecCommandApprovalParams {
     pub cwd: PathBuf,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub reason: Option<String>,
+    pub parsed_cmd: Vec<ParsedCommand>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 pub struct ExecCommandApprovalResponse {
     pub decision: ReviewDecision,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 pub struct ApplyPatchApprovalResponse {
     pub decision: ReviewDecision,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(rename_all = "camelCase")]
 pub struct FuzzyFileSearchParams {
@@ -721,30 +752,32 @@ pub struct FuzzyFileSearchParams {
 }
 
 /// Superset of [`codex_file_search::FileMatch`]
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 pub struct FuzzyFileSearchResult {
     pub root: String,
     pub path: String,
+    pub file_name: String,
     pub score: u32,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub indices: Option<Vec<u32>>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 pub struct FuzzyFileSearchResponse {
     pub files: Vec<FuzzyFileSearchResult>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct LoginChatGptCompleteNotification {
+    #[schemars(with = "String")]
     pub login_id: Uuid,
     pub success: bool,
     #[serde(skip_serializing_if = "Option::is_none")]
     pub error: Option<String>,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct SessionConfiguredNotification {
     /// Name left as session_id instead of conversation_id for backwards compatibility.
@@ -772,7 +805,7 @@ pub struct SessionConfiguredNotification {
     pub rollout_path: PathBuf,
 }
 
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 pub struct AuthStatusChangeNotification {
     /// Current authentication method; omitted if signed out.
@@ -781,7 +814,7 @@ pub struct AuthStatusChangeNotification {
 }
 
 /// Notification sent from the server to the client.
-#[derive(Serialize, Deserialize, Debug, Clone, TS, Display)]
+#[derive(Serialize, Deserialize, Debug, Clone, JsonSchema, TS, Display)]
 #[serde(tag = "method", content = "params", rename_all = "camelCase")]
 #[strum(serialize_all = "camelCase")]
 pub enum ServerNotification {
@@ -814,7 +847,7 @@ impl TryFrom<JSONRPCNotification> for ServerNotification {
 }
 
 /// Notification sent from the client to the server.
-#[derive(Serialize, Deserialize, Debug, Clone, TS, Display)]
+#[derive(Serialize, Deserialize, Debug, Clone, JsonSchema, TS, Display)]
 #[serde(tag = "method", content = "params", rename_all = "camelCase")]
 #[strum(serialize_all = "camelCase")]
 pub enum ClientNotification {
@@ -903,6 +936,9 @@ mod tests {
             command: vec!["echo".to_string(), "hello".to_string()],
             cwd: PathBuf::from("/tmp"),
             reason: Some("because tests".to_string()),
+            parsed_cmd: vec![ParsedCommand::Unknown {
+                cmd: "echo hello".to_string(),
+            }],
         };
         let request = ServerRequest::ExecCommandApproval {
             request_id: RequestId::Integer(7),
@@ -919,6 +955,12 @@ mod tests {
                     "command": ["echo", "hello"],
                     "cwd": "/tmp",
                     "reason": "because tests",
+                    "parsedCmd": [
+                        {
+                            "type": "unknown",
+                            "cmd": "echo hello"
+                        }
+                    ]
                 }
             }),
             serde_json::to_value(&request)?,

codex-rs/app-server/README.md

@@ -0,0 +1,15 @@
+# codex-app-server
+
+`codex app-server` is the harness Codex uses to power rich interfaces such as the [Codex VS Code extension](https://marketplace.visualstudio.com/items?itemName=openai.chatgpt). The message schema is currently unstable, but those who wish to build experimental UIs on top of Codex may find it valuable.
+
+## Protocol
+
+Similar to [MCP](https://modelcontextprotocol.io/), `codex app-server` supports bidirectional communication, streaming JSONL over stdio. The protocol is JSON-RPC 2.0, though the `"jsonrpc":"2.0"` header is omitted.
+
+## Message Schema
+
+Currently, you can dump a TypeScript version of the schema using `codex generate-ts`. It is specific to the version of Codex you used to run `generate-ts`, so the two are guaranteed to be compatible.
+
+```
+codex generate-ts --out DIR
+```

codex-rs/app-server/src/codex_message_processor.rs

@@ -53,6 +53,7 @@ use codex_core::AuthManager;
 use codex_core::CodexConversation;
 use codex_core::ConversationManager;
 use codex_core::Cursor as RolloutCursor;
+use codex_core::INTERACTIVE_SESSION_SOURCES;
 use codex_core::NewConversation;
 use codex_core::RolloutRecorder;
 use codex_core::SessionMeta;
@@ -83,6 +84,7 @@ use codex_login::ServerOptions as LoginServerOptions;
 use codex_login::ShutdownHandle;
 use codex_login::run_login_server;
 use codex_protocol::ConversationId;
+use codex_protocol::config_types::ForcedLoginMethod;
 use codex_protocol::models::ContentItem;
 use codex_protocol::models::ResponseItem;
 use codex_protocol::protocol::InputMessageKind;
@@ -242,6 +244,19 @@ impl CodexMessageProcessor {
     }
 
     async fn login_api_key(&mut self, request_id: RequestId, params: LoginApiKeyParams) {
+        if matches!(
+            self.config.forced_login_method,
+            Some(ForcedLoginMethod::Chatgpt)
+        ) {
+            let error = JSONRPCErrorError {
+                code: INVALID_REQUEST_ERROR_CODE,
+                message: "API key login is disabled. Use ChatGPT login instead.".to_string(),
+                data: None,
+            };
+            self.outgoing.send_error(request_id, error).await;
+            return;
+        }
+
         {
             let mut guard = self.active_login.lock().await;
             if let Some(active) = guard.take() {
@@ -277,9 +292,23 @@ impl CodexMessageProcessor {
     async fn login_chatgpt(&mut self, request_id: RequestId) {
         let config = self.config.as_ref();
 
+        if matches!(config.forced_login_method, Some(ForcedLoginMethod::Api)) {
+            let error = JSONRPCErrorError {
+                code: INVALID_REQUEST_ERROR_CODE,
+                message: "ChatGPT login is disabled. Use API key login instead.".to_string(),
+                data: None,
+            };
+            self.outgoing.send_error(request_id, error).await;
+            return;
+        }
+
         let opts = LoginServerOptions {
             open_browser: false,
-            ..LoginServerOptions::new(config.codex_home.clone(), CLIENT_ID.to_string())
+            ..LoginServerOptions::new(
+                config.codex_home.clone(),
+                CLIENT_ID.to_string(),
+                config.forced_chatgpt_workspace_id.clone(),
+            )
         };
 
         enum LoginChatGptReply {
@@ -499,7 +528,7 @@ impl CodexMessageProcessor {
     }
 
     async fn get_user_saved_config(&self, request_id: RequestId) {
-        let toml_value = match load_config_as_toml(&self.config.codex_home) {
+        let toml_value = match load_config_as_toml(&self.config.codex_home).await {
             Ok(val) => val,
             Err(err) => {
                 let error = JSONRPCErrorError {
@@ -602,6 +631,7 @@ impl CodexMessageProcessor {
             env,
             with_escalated_permissions: None,
             justification: None,
+            arg0: None,
         };
 
         let effective_policy = params
@@ -652,18 +682,19 @@ impl CodexMessageProcessor {
     }
 
     async fn process_new_conversation(&self, request_id: RequestId, params: NewConversationParams) {
-        let config = match derive_config_from_params(params, self.codex_linux_sandbox_exe.clone()) {
-            Ok(config) => config,
-            Err(err) => {
-                let error = JSONRPCErrorError {
-                    code: INVALID_REQUEST_ERROR_CODE,
-                    message: format!("error deriving config: {err}"),
-                    data: None,
-                };
-                self.outgoing.send_error(request_id, error).await;
-                return;
-            }
-        };
+        let config =
+            match derive_config_from_params(params, self.codex_linux_sandbox_exe.clone()).await {
+                Ok(config) => config,
+                Err(err) => {
+                    let error = JSONRPCErrorError {
+                        code: INVALID_REQUEST_ERROR_CODE,
+                        message: format!("error deriving config: {err}"),
+                        data: None,
+                    };
+                    self.outgoing.send_error(request_id, error).await;
+                    return;
+                }
+            };
 
         match self.conversation_manager.new_conversation(config).await {
             Ok(conversation_id) => {
@@ -708,6 +739,7 @@ impl CodexMessageProcessor {
             &self.config.codex_home,
             page_size,
             cursor_ref,
+            INTERACTIVE_SESSION_SOURCES,
         )
         .await
         {
@@ -750,7 +782,7 @@ impl CodexMessageProcessor {
         // Derive a Config using the same logic as new conversation, honoring overrides if provided.
         let config = match params.overrides {
             Some(overrides) => {
-                derive_config_from_params(overrides, self.codex_linux_sandbox_exe.clone())
+                derive_config_from_params(overrides, self.codex_linux_sandbox_exe.clone()).await
             }
             None => Ok(self.config.as_ref().clone()),
         };
@@ -1281,6 +1313,7 @@ async fn apply_bespoke_event_handling(
             command,
             cwd,
             reason,
+            parsed_cmd,
         }) => {
             let params = ExecCommandApprovalParams {
                 conversation_id,
@@ -1288,6 +1321,7 @@ async fn apply_bespoke_event_handling(
                 command,
                 cwd,
                 reason,
+                parsed_cmd,
             };
             let rx = outgoing
                 .send_request(ServerRequestPayload::ExecCommandApproval(params))
@@ -1318,7 +1352,7 @@ async fn apply_bespoke_event_handling(
     }
 }
 
-fn derive_config_from_params(
+async fn derive_config_from_params(
     params: NewConversationParams,
     codex_linux_sandbox_exe: Option<PathBuf>,
 ) -> std::io::Result<Config> {
@@ -1348,6 +1382,7 @@ fn derive_config_from_params(
         include_view_image_tool: None,
         show_raw_agent_reasoning: None,
         tools_web_search_request: None,
+        additional_writable_roots: Vec::new(),
     };
 
     let cli_overrides = cli_overrides
@@ -1356,7 +1391,7 @@ fn derive_config_from_params(
         .map(|(k, v)| (k, json_to_toml(v)))
         .collect();
 
-    Config::load_with_cli_overrides(cli_overrides, overrides)
+    Config::load_with_cli_overrides(cli_overrides, overrides).await
 }
 
 async fn on_patch_approval_response(

codex-rs/app-server/src/fuzzy_file_search.rs

@@ -1,5 +1,6 @@
 use std::num::NonZero;
 use std::num::NonZeroUsize;
+use std::path::Path;
 use std::path::PathBuf;
 use std::sync::Arc;
 use std::sync::atomic::AtomicBool;
@@ -56,9 +57,16 @@ pub(crate) async fn run_fuzzy_file_search(
         match res {
             Ok(Ok((root, res))) => {
                 for m in res.matches {
+                    let path = m.path;
+                    //TODO(shijie): Move file name generation to file_search lib.
+                    let file_name = Path::new(&path)
+                        .file_name()
+                        .map(|name| name.to_string_lossy().into_owned())
+                        .unwrap_or_else(|| path.clone());
                     let result = FuzzyFileSearchResult {
                         root: root.clone(),
-                        path: m.path,
+                        path,
+                        file_name,
                         score: m.score,
                         indices: m.indices,
                     };

codex-rs/app-server/src/lib.rs

@@ -81,6 +81,7 @@ pub async fn run_main(
         )
     })?;
     let config = Config::load_with_cli_overrides(cli_kv_overrides, ConfigOverrides::default())
+        .await
         .map_err(|e| {
             std::io::Error::new(ErrorKind::InvalidData, format!("error loading config: {e}"))
         })?;

codex-rs/app-server/src/message_processor.rs

@@ -17,6 +17,7 @@ use codex_core::ConversationManager;
 use codex_core::config::Config;
 use codex_core::default_client::USER_AGENT_SUFFIX;
 use codex_core::default_client::get_codex_user_agent;
+use codex_protocol::protocol::SessionSource;
 use std::sync::Arc;
 
 pub(crate) struct MessageProcessor {
@@ -34,8 +35,11 @@ impl MessageProcessor {
         config: Arc<Config>,
     ) -> Self {
         let outgoing = Arc::new(outgoing);
-        let auth_manager = AuthManager::shared(config.codex_home.clone());
-        let conversation_manager = Arc::new(ConversationManager::new(auth_manager.clone()));
+        let auth_manager = AuthManager::shared(config.codex_home.clone(), false);
+        let conversation_manager = Arc::new(ConversationManager::new(
+            auth_manager.clone(),
+            SessionSource::VSCode,
+        ));
         let codex_message_processor = CodexMessageProcessor::new(
             auth_manager,
             conversation_manager,

codex-rs/app-server/tests/suite/auth.rs

@@ -5,6 +5,7 @@ use app_test_support::to_response;
 use codex_app_server_protocol::AuthMode;
 use codex_app_server_protocol::GetAuthStatusParams;
 use codex_app_server_protocol::GetAuthStatusResponse;
+use codex_app_server_protocol::JSONRPCError;
 use codex_app_server_protocol::JSONRPCResponse;
 use codex_app_server_protocol::LoginApiKeyParams;
 use codex_app_server_protocol::LoginApiKeyResponse;
@@ -57,6 +58,19 @@ sandbox_mode = "danger-full-access"
     )
 }
 
+fn create_config_toml_forced_login(codex_home: &Path, forced_method: &str) -> std::io::Result<()> {
+    let config_toml = codex_home.join("config.toml");
+    let contents = format!(
+        r#"
+model = "mock-model"
+approval_policy = "never"
+sandbox_mode = "danger-full-access"
+forced_login_method = "{forced_method}"
+"#
+    );
+    std::fs::write(config_toml, contents)
+}
+
 async fn login_with_api_key_via_request(mcp: &mut McpProcess, api_key: &str) {
     let request_id = mcp
         .send_login_api_key_request(LoginApiKeyParams {
@@ -221,3 +235,38 @@ async fn get_auth_status_with_api_key_no_include_token() {
     assert_eq!(status.auth_method, Some(AuthMode::ApiKey));
     assert!(status.auth_token.is_none(), "token must be omitted");
 }
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn login_api_key_rejected_when_forced_chatgpt() {
+    let codex_home = TempDir::new().unwrap_or_else(|e| panic!("create tempdir: {e}"));
+    create_config_toml_forced_login(codex_home.path(), "chatgpt")
+        .unwrap_or_else(|err| panic!("write config.toml: {err}"));
+
+    let mut mcp = McpProcess::new(codex_home.path())
+        .await
+        .expect("spawn mcp process");
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize())
+        .await
+        .expect("init timeout")
+        .expect("init failed");
+
+    let request_id = mcp
+        .send_login_api_key_request(LoginApiKeyParams {
+            api_key: "sk-test-key".to_string(),
+        })
+        .await
+        .expect("send loginApiKey");
+
+    let err: JSONRPCError = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_error_message(RequestId::Integer(request_id)),
+    )
+    .await
+    .expect("loginApiKey error timeout")
+    .expect("loginApiKey error");
+
+    assert_eq!(
+        err.error.message,
+        "API key login is disabled. Use ChatGPT login instead."
+    );
+}

codex-rs/app-server/tests/suite/codex_message_processor_flow.rs

@@ -27,6 +27,7 @@ use codex_core::protocol_config_types::ReasoningEffort;
 use codex_core::protocol_config_types::ReasoningSummary;
 use codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
 use codex_protocol::config_types::SandboxMode;
+use codex_protocol::parse_command::ParsedCommand;
 use codex_protocol::protocol::Event;
 use codex_protocol::protocol::EventMsg;
 use codex_protocol::protocol::InputMessageKind;
@@ -311,6 +312,9 @@ async fn test_send_user_turn_changes_approval_policy_behavior() {
             ],
             cwd: working_directory.clone(),
             reason: None,
+            parsed_cmd: vec![ParsedCommand::Unknown {
+                cmd: "python3 -c 'print(42)'".to_string()
+            }],
         },
         params
     );

codex-rs/app-server/tests/suite/config.rs

@@ -11,6 +11,7 @@ use codex_app_server_protocol::SandboxSettings;
 use codex_app_server_protocol::Tools;
 use codex_app_server_protocol::UserSavedConfig;
 use codex_core::protocol::AskForApproval;
+use codex_protocol::config_types::ForcedLoginMethod;
 use codex_protocol::config_types::ReasoningEffort;
 use codex_protocol::config_types::ReasoningSummary;
 use codex_protocol::config_types::SandboxMode;
@@ -33,6 +34,8 @@ model_reasoning_summary = "detailed"
 model_reasoning_effort = "high"
 model_verbosity = "medium"
 profile = "test"
+forced_chatgpt_workspace_id = "12345678-0000-0000-0000-000000000000"
+forced_login_method = "chatgpt"
 
 [sandbox_workspace_write]
 writable_roots = ["/tmp"]
@@ -92,6 +95,8 @@ async fn get_config_toml_parses_all_fields() {
                 exclude_tmpdir_env_var: Some(true),
                 exclude_slash_tmp: Some(true),
             }),
+            forced_chatgpt_workspace_id: Some("12345678-0000-0000-0000-000000000000".into()),
+            forced_login_method: Some(ForcedLoginMethod::Chatgpt),
             model: Some("gpt-5-codex".into()),
             model_reasoning_effort: Some(ReasoningEffort::High),
             model_reasoning_summary: Some(ReasoningSummary::Detailed),
@@ -149,6 +154,8 @@ async fn get_config_toml_empty() {
             approval_policy: None,
             sandbox_mode: None,
             sandbox_settings: None,
+            forced_chatgpt_workspace_id: None,
+            forced_login_method: None,
             model: None,
             model_reasoning_effort: None,
             model_reasoning_summary: None,

codex-rs/app-server/tests/suite/fuzzy_file_search.rs

@@ -1,3 +1,5 @@
+use anyhow::Context;
+use anyhow::Result;
 use app_test_support::McpProcess;
 use codex_app_server_protocol::JSONRPCResponse;
 use codex_app_server_protocol::RequestId;
@@ -9,30 +11,41 @@ use tokio::time::timeout;
 const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn test_fuzzy_file_search_sorts_and_includes_indices() {
+async fn test_fuzzy_file_search_sorts_and_includes_indices() -> Result<()> {
     // Prepare a temporary Codex home and a separate root with test files.
-    let codex_home = TempDir::new().expect("create temp codex home");
-    let root = TempDir::new().expect("create temp search root");
+    let codex_home = TempDir::new().context("create temp codex home")?;
+    let root = TempDir::new().context("create temp search root")?;
 
-    // Create files designed to have deterministic ordering for query "abc".
-    std::fs::write(root.path().join("abc"), "x").expect("write file abc");
-    std::fs::write(root.path().join("abcde"), "x").expect("write file abcx");
-    std::fs::write(root.path().join("abexy"), "x").expect("write file abcx");
-    std::fs::write(root.path().join("zzz.txt"), "x").expect("write file zzz");
+    // Create files designed to have deterministic ordering for query "abe".
+    std::fs::write(root.path().join("abc"), "x").context("write file abc")?;
+    std::fs::write(root.path().join("abcde"), "x").context("write file abcde")?;
+    std::fs::write(root.path().join("abexy"), "x").context("write file abexy")?;
+    std::fs::write(root.path().join("zzz.txt"), "x").context("write file zzz")?;
+    let sub_dir = root.path().join("sub");
+    std::fs::create_dir_all(&sub_dir).context("create sub dir")?;
+    let sub_abce_path = sub_dir.join("abce");
+    std::fs::write(&sub_abce_path, "x").context("write file sub/abce")?;
+    let sub_abce_rel = sub_abce_path
+        .strip_prefix(root.path())
+        .context("strip root prefix from sub/abce")?
+        .to_string_lossy()
+        .to_string();
 
     // Start MCP server and initialize.
-    let mut mcp = McpProcess::new(codex_home.path()).await.expect("spawn mcp");
+    let mut mcp = McpProcess::new(codex_home.path())
+        .await
+        .context("spawn mcp")?;
     timeout(DEFAULT_READ_TIMEOUT, mcp.initialize())
         .await
-        .expect("init timeout")
-        .expect("init failed");
+        .context("init timeout")?
+        .context("init failed")?;
 
     let root_path = root.path().to_string_lossy().to_string();
     // Send fuzzyFileSearch request.
     let request_id = mcp
         .send_fuzzy_file_search_request("abe", vec![root_path.clone()], None)
         .await
-        .expect("send fuzzyFileSearch");
+        .context("send fuzzyFileSearch")?;
 
     // Read response and verify shape and ordering.
     let resp: JSONRPCResponse = timeout(
@@ -40,39 +53,65 @@ async fn test_fuzzy_file_search_sorts_and_includes_indices() {
         mcp.read_stream_until_response_message(RequestId::Integer(request_id)),
     )
     .await
-    .expect("fuzzyFileSearch timeout")
-    .expect("fuzzyFileSearch resp");
+    .context("fuzzyFileSearch timeout")?
+    .context("fuzzyFileSearch resp")?;
 
     let value = resp.result;
+    // The path separator on Windows affects the score.
+    let expected_score = if cfg!(windows) { 69 } else { 72 };
+
     assert_eq!(
         value,
         json!({
             "files": [
-                { "root": root_path.clone(), "path": "abexy", "score": 88, "indices": [0, 1, 2] },
-                { "root": root_path.clone(), "path": "abcde", "score": 74, "indices": [0, 1, 4] },
+                {
+                    "root": root_path.clone(),
+                    "path": "abexy",
+                    "file_name": "abexy",
+                    "score": 88,
+                    "indices": [0, 1, 2],
+                },
+                {
+                    "root": root_path.clone(),
+                    "path": "abcde",
+                    "file_name": "abcde",
+                    "score": 74,
+                    "indices": [0, 1, 4],
+                },
+                {
+                    "root": root_path.clone(),
+                    "path": sub_abce_rel,
+                    "file_name": "abce",
+                    "score": expected_score,
+                    "indices": [4, 5, 7],
+                },
             ]
         })
     );
+
+    Ok(())
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn test_fuzzy_file_search_accepts_cancellation_token() {
-    let codex_home = TempDir::new().expect("create temp codex home");
-    let root = TempDir::new().expect("create temp search root");
+async fn test_fuzzy_file_search_accepts_cancellation_token() -> Result<()> {
+    let codex_home = TempDir::new().context("create temp codex home")?;
+    let root = TempDir::new().context("create temp search root")?;
 
-    std::fs::write(root.path().join("alpha.txt"), "contents").expect("write alpha");
+    std::fs::write(root.path().join("alpha.txt"), "contents").context("write alpha")?;
 
-    let mut mcp = McpProcess::new(codex_home.path()).await.expect("spawn mcp");
+    let mut mcp = McpProcess::new(codex_home.path())
+        .await
+        .context("spawn mcp")?;
     timeout(DEFAULT_READ_TIMEOUT, mcp.initialize())
         .await
-        .expect("init timeout")
-        .expect("init failed");
+        .context("init timeout")?
+        .context("init failed")?;
 
     let root_path = root.path().to_string_lossy().to_string();
     let request_id = mcp
         .send_fuzzy_file_search_request("alp", vec![root_path.clone()], None)
         .await
-        .expect("send fuzzyFileSearch");
+        .context("send fuzzyFileSearch")?;
 
     let request_id_2 = mcp
         .send_fuzzy_file_search_request(
@@ -81,24 +120,27 @@ async fn test_fuzzy_file_search_accepts_cancellation_token() {
             Some(request_id.to_string()),
         )
         .await
-        .expect("send fuzzyFileSearch");
+        .context("send fuzzyFileSearch")?;
 
     let resp: JSONRPCResponse = timeout(
         DEFAULT_READ_TIMEOUT,
         mcp.read_stream_until_response_message(RequestId::Integer(request_id_2)),
     )
     .await
-    .expect("fuzzyFileSearch timeout")
-    .expect("fuzzyFileSearch resp");
+    .context("fuzzyFileSearch timeout")?
+    .context("fuzzyFileSearch resp")?;
 
     let files = resp
         .result
         .get("files")
-        .and_then(|value| value.as_array())
-        .cloned()
-        .expect("files array");
+        .context("files key missing")?
+        .as_array()
+        .context("files not array")?
+        .clone();
 
     assert_eq!(files.len(), 1);
     assert_eq!(files[0]["root"], root_path);
     assert_eq!(files[0]["path"], "alpha.txt");
+
+    Ok(())
 }

codex-rs/app-server/tests/suite/login.rs

@@ -7,6 +7,7 @@ use codex_app_server_protocol::CancelLoginChatGptParams;
 use codex_app_server_protocol::CancelLoginChatGptResponse;
 use codex_app_server_protocol::GetAuthStatusParams;
 use codex_app_server_protocol::GetAuthStatusResponse;
+use codex_app_server_protocol::JSONRPCError;
 use codex_app_server_protocol::JSONRPCResponse;
 use codex_app_server_protocol::LoginChatGptResponse;
 use codex_app_server_protocol::LogoutChatGptResponse;
@@ -144,3 +145,97 @@ async fn login_and_cancel_chatgpt() {
         eprintln!("warning: did not observe login_chat_gpt_complete notification after cancel");
     }
 }
+
+fn create_config_toml_forced_login(codex_home: &Path, forced_method: &str) -> std::io::Result<()> {
+    let config_toml = codex_home.join("config.toml");
+    let contents = format!(
+        r#"
+model = "mock-model"
+approval_policy = "never"
+sandbox_mode = "danger-full-access"
+forced_login_method = "{forced_method}"
+"#
+    );
+    std::fs::write(config_toml, contents)
+}
+
+fn create_config_toml_forced_workspace(
+    codex_home: &Path,
+    workspace_id: &str,
+) -> std::io::Result<()> {
+    let config_toml = codex_home.join("config.toml");
+    let contents = format!(
+        r#"
+model = "mock-model"
+approval_policy = "never"
+sandbox_mode = "danger-full-access"
+forced_chatgpt_workspace_id = "{workspace_id}"
+"#
+    );
+    std::fs::write(config_toml, contents)
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn login_chatgpt_rejected_when_forced_api() {
+    let codex_home = TempDir::new().unwrap_or_else(|e| panic!("create tempdir: {e}"));
+    create_config_toml_forced_login(codex_home.path(), "api")
+        .unwrap_or_else(|err| panic!("write config.toml: {err}"));
+
+    let mut mcp = McpProcess::new(codex_home.path())
+        .await
+        .expect("spawn mcp process");
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize())
+        .await
+        .expect("init timeout")
+        .expect("init failed");
+
+    let request_id = mcp
+        .send_login_chat_gpt_request()
+        .await
+        .expect("send loginChatGpt");
+    let err: JSONRPCError = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_error_message(RequestId::Integer(request_id)),
+    )
+    .await
+    .expect("loginChatGpt error timeout")
+    .expect("loginChatGpt error");
+
+    assert_eq!(
+        err.error.message,
+        "ChatGPT login is disabled. Use API key login instead."
+    );
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn login_chatgpt_includes_forced_workspace_query_param() {
+    let codex_home = TempDir::new().unwrap_or_else(|e| panic!("create tempdir: {e}"));
+    create_config_toml_forced_workspace(codex_home.path(), "ws-forced")
+        .unwrap_or_else(|err| panic!("write config.toml: {err}"));
+
+    let mut mcp = McpProcess::new(codex_home.path())
+        .await
+        .expect("spawn mcp process");
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize())
+        .await
+        .expect("init timeout")
+        .expect("init failed");
+
+    let request_id = mcp
+        .send_login_chat_gpt_request()
+        .await
+        .expect("send loginChatGpt");
+    let resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(request_id)),
+    )
+    .await
+    .expect("loginChatGpt timeout")
+    .expect("loginChatGpt response");
+
+    let login: LoginChatGptResponse = to_response(resp).expect("deserialize login resp");
+    assert!(
+        login.auth_url.contains("allowed_workspace_id=ws-forced"),
+        "auth URL should include forced workspace"
+    );
+}

codex-rs/apply-patch/Cargo.toml

@@ -23,5 +23,6 @@ tree-sitter-bash = { workspace = true }
 
 [dev-dependencies]
 assert_cmd = { workspace = true }
+assert_matches = { workspace = true }
 pretty_assertions = { workspace = true }
 tempfile = { workspace = true }

codex-rs/apply-patch/src/lib.rs

@@ -843,6 +843,7 @@ pub fn print_summary(
 #[cfg(test)]
 mod tests {
     use super::*;
+    use assert_matches::assert_matches;
     use pretty_assertions::assert_eq;
     use std::fs;
     use std::string::ToString;
@@ -894,10 +895,10 @@ mod tests {
 
     fn assert_not_match(script: &str) {
         let args = args_bash(script);
-        assert!(matches!(
+        assert_matches!(
             maybe_parse_apply_patch(&args),
             MaybeApplyPatch::NotApplyPatch
-        ));
+        );
     }
 
     #[test]
@@ -905,10 +906,10 @@ mod tests {
         let patch = "*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch".to_string();
         let args = vec![patch];
         let dir = tempdir().unwrap();
-        assert!(matches!(
+        assert_matches!(
             maybe_parse_apply_patch_verified(&args, dir.path()),
             MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation)
-        ));
+        );
     }
 
     #[test]
@@ -916,10 +917,10 @@ mod tests {
         let script = "*** Begin Patch\n*** Add File: foo\n+hi\n*** End Patch";
         let args = args_bash(script);
         let dir = tempdir().unwrap();
-        assert!(matches!(
+        assert_matches!(
             maybe_parse_apply_patch_verified(&args, dir.path()),
             MaybeApplyPatchVerified::CorrectnessError(ApplyPatchError::ImplicitInvocation)
-        ));
+        );
     }
 
     #[test]

codex-rs/async-utils/Cargo.toml

@@ -0,0 +1,15 @@
+[package]
+edition.workspace = true
+name = "codex-async-utils"
+version.workspace = true
+
+[lints]
+workspace = true
+
+[dependencies]
+async-trait.workspace = true
+tokio = { workspace = true, features = ["macros", "rt", "rt-multi-thread", "time"] }
+tokio-util.workspace = true
+
+[dev-dependencies]
+pretty_assertions.workspace = true

codex-rs/async-utils/src/lib.rs

@@ -0,0 +1,86 @@
+use async_trait::async_trait;
+use std::future::Future;
+use tokio_util::sync::CancellationToken;
+
+#[derive(Debug, PartialEq, Eq)]
+pub enum CancelErr {
+    Cancelled,
+}
+
+#[async_trait]
+pub trait OrCancelExt: Sized {
+    type Output;
+
+    async fn or_cancel(self, token: &CancellationToken) -> Result<Self::Output, CancelErr>;
+}
+
+#[async_trait]
+impl<F> OrCancelExt for F
+where
+    F: Future + Send,
+    F::Output: Send,
+{
+    type Output = F::Output;
+
+    async fn or_cancel(self, token: &CancellationToken) -> Result<Self::Output, CancelErr> {
+        tokio::select! {
+            _ = token.cancelled() => Err(CancelErr::Cancelled),
+            res = self => Ok(res),
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use pretty_assertions::assert_eq;
+    use std::time::Duration;
+    use tokio::task;
+    use tokio::time::sleep;
+
+    #[tokio::test]
+    async fn returns_ok_when_future_completes_first() {
+        let token = CancellationToken::new();
+        let value = async { 42 };
+
+        let result = value.or_cancel(&token).await;
+
+        assert_eq!(Ok(42), result);
+    }
+
+    #[tokio::test]
+    async fn returns_err_when_token_cancelled_first() {
+        let token = CancellationToken::new();
+        let token_clone = token.clone();
+
+        let cancel_handle = task::spawn(async move {
+            sleep(Duration::from_millis(10)).await;
+            token_clone.cancel();
+        });
+
+        let result = async {
+            sleep(Duration::from_millis(100)).await;
+            7
+        }
+        .or_cancel(&token)
+        .await;
+
+        cancel_handle.await.expect("cancel task panicked");
+        assert_eq!(Err(CancelErr::Cancelled), result);
+    }
+
+    #[tokio::test]
+    async fn returns_err_when_token_already_cancelled() {
+        let token = CancellationToken::new();
+        token.cancel();
+
+        let result = async {
+            sleep(Duration::from_millis(50)).await;
+            5
+        }
+        .or_cancel(&token)
+        .await;
+
+        assert_eq!(Err(CancelErr::Cancelled), result);
+    }
+}

codex-rs/chatgpt/src/apply_command.rs

@@ -29,7 +29,8 @@ pub async fn run_apply_command(
             .parse_overrides()
             .map_err(anyhow::Error::msg)?,
         ConfigOverrides::default(),
-    )?;
+    )
+    .await?;
 
     init_chatgpt_token_from_auth(&config.codex_home).await?;
 

codex-rs/cli/Cargo.toml

@@ -19,8 +19,10 @@ anyhow = { workspace = true }
 clap = { workspace = true, features = ["derive"] }
 clap_complete = { workspace = true }
 codex-app-server = { workspace = true }
+codex-app-server-protocol = { workspace = true }
 codex-arg0 = { workspace = true }
 codex-chatgpt = { workspace = true }
+codex-cloud-tasks = { path = "../cloud-tasks" }
 codex-common = { workspace = true, features = ["cli"] }
 codex-core = { workspace = true }
 codex-exec = { workspace = true }
@@ -28,11 +30,11 @@ codex-login = { workspace = true }
 codex-mcp-server = { workspace = true }
 codex-process-hardening = { workspace = true }
 codex-protocol = { workspace = true }
-codex-app-server-protocol = { workspace = true }
 codex-protocol-ts = { workspace = true }
 codex-responses-api-proxy = { workspace = true }
+codex-rmcp-client = { workspace = true }
+codex-stdio-to-uds = { workspace = true }
 codex-tui = { workspace = true }
-codex-cloud-tasks = { path = "../cloud-tasks" }
 ctor = { workspace = true }
 owo-colors = { workspace = true }
 serde_json = { workspace = true }
@@ -47,6 +49,7 @@ tokio = { workspace = true, features = [
 
 [dev-dependencies]
 assert_cmd = { workspace = true }
+assert_matches = { workspace = true }
 predicates = { workspace = true }
 pretty_assertions = { workspace = true }
 tempfile = { workspace = true }

codex-rs/cli/src/debug_sandbox.rs

@@ -73,7 +73,8 @@ async fn run_command_under_sandbox(
             codex_linux_sandbox_exe,
             ..Default::default()
         },
-    )?;
+    )
+    .await?;
 
     // In practice, this should be `std::env::current_dir()` because this CLI
     // does not support `--cwd`, but let's use the config value for consistency.

codex-rs/cli/src/login.rs

@@ -9,10 +9,20 @@ use codex_core::config::ConfigOverrides;
 use codex_login::ServerOptions;
 use codex_login::run_device_code_login;
 use codex_login::run_login_server;
+use codex_protocol::config_types::ForcedLoginMethod;
+use std::io::IsTerminal;
+use std::io::Read;
 use std::path::PathBuf;
 
-pub async fn login_with_chatgpt(codex_home: PathBuf) -> std::io::Result<()> {
-    let opts = ServerOptions::new(codex_home, CLIENT_ID.to_string());
+pub async fn login_with_chatgpt(
+    codex_home: PathBuf,
+    forced_chatgpt_workspace_id: Option<String>,
+) -> std::io::Result<()> {
+    let opts = ServerOptions::new(
+        codex_home,
+        CLIENT_ID.to_string(),
+        forced_chatgpt_workspace_id,
+    );
     let server = run_login_server(opts)?;
 
     eprintln!(
@@ -24,9 +34,16 @@ pub async fn login_with_chatgpt(codex_home: PathBuf) -> std::io::Result<()> {
 }
 
 pub async fn run_login_with_chatgpt(cli_config_overrides: CliConfigOverrides) -> ! {
-    let config = load_config_or_exit(cli_config_overrides);
+    let config = load_config_or_exit(cli_config_overrides).await;
 
-    match login_with_chatgpt(config.codex_home).await {
+    if matches!(config.forced_login_method, Some(ForcedLoginMethod::Api)) {
+        eprintln!("ChatGPT login is disabled. Use API key login instead.");
+        std::process::exit(1);
+    }
+
+    let forced_chatgpt_workspace_id = config.forced_chatgpt_workspace_id.clone();
+
+    match login_with_chatgpt(config.codex_home, forced_chatgpt_workspace_id).await {
         Ok(_) => {
             eprintln!("Successfully logged in");
             std::process::exit(0);
@@ -42,7 +59,12 @@ pub async fn run_login_with_api_key(
     cli_config_overrides: CliConfigOverrides,
     api_key: String,
 ) -> ! {
-    let config = load_config_or_exit(cli_config_overrides);
+    let config = load_config_or_exit(cli_config_overrides).await;
+
+    if matches!(config.forced_login_method, Some(ForcedLoginMethod::Chatgpt)) {
+        eprintln!("API key login is disabled. Use ChatGPT login instead.");
+        std::process::exit(1);
+    }
 
     match login_with_api_key(&config.codex_home, &api_key) {
         Ok(_) => {
@@ -56,16 +78,49 @@ pub async fn run_login_with_api_key(
     }
 }
 
+pub fn read_api_key_from_stdin() -> String {
+    let mut stdin = std::io::stdin();
+
+    if stdin.is_terminal() {
+        eprintln!(
+            "--with-api-key expects the API key on stdin. Try piping it, e.g. `printenv OPENAI_API_KEY | codex login --with-api-key`."
+        );
+        std::process::exit(1);
+    }
+
+    eprintln!("Reading API key from stdin...");
+
+    let mut buffer = String::new();
+    if let Err(err) = stdin.read_to_string(&mut buffer) {
+        eprintln!("Failed to read API key from stdin: {err}");
+        std::process::exit(1);
+    }
+
+    let api_key = buffer.trim().to_string();
+    if api_key.is_empty() {
+        eprintln!("No API key provided via stdin.");
+        std::process::exit(1);
+    }
+
+    api_key
+}
+
 /// Login using the OAuth device code flow.
 pub async fn run_login_with_device_code(
     cli_config_overrides: CliConfigOverrides,
     issuer_base_url: Option<String>,
     client_id: Option<String>,
 ) -> ! {
-    let config = load_config_or_exit(cli_config_overrides);
+    let config = load_config_or_exit(cli_config_overrides).await;
+    if matches!(config.forced_login_method, Some(ForcedLoginMethod::Api)) {
+        eprintln!("ChatGPT login is disabled. Use API key login instead.");
+        std::process::exit(1);
+    }
+    let forced_chatgpt_workspace_id = config.forced_chatgpt_workspace_id.clone();
     let mut opts = ServerOptions::new(
         config.codex_home,
         client_id.unwrap_or(CLIENT_ID.to_string()),
+        forced_chatgpt_workspace_id,
     );
     if let Some(iss) = issuer_base_url {
         opts.issuer = iss;
@@ -83,7 +138,7 @@ pub async fn run_login_with_device_code(
 }
 
 pub async fn run_login_status(cli_config_overrides: CliConfigOverrides) -> ! {
-    let config = load_config_or_exit(cli_config_overrides);
+    let config = load_config_or_exit(cli_config_overrides).await;
 
     match CodexAuth::from_codex_home(&config.codex_home) {
         Ok(Some(auth)) => match auth.mode {
@@ -114,7 +169,7 @@ pub async fn run_login_status(cli_config_overrides: CliConfigOverrides) -> ! {
 }
 
 pub async fn run_logout(cli_config_overrides: CliConfigOverrides) -> ! {
-    let config = load_config_or_exit(cli_config_overrides);
+    let config = load_config_or_exit(cli_config_overrides).await;
 
     match logout(&config.codex_home) {
         Ok(true) => {
@@ -132,7 +187,7 @@ pub async fn run_logout(cli_config_overrides: CliConfigOverrides) -> ! {
     }
 }
 
-fn load_config_or_exit(cli_config_overrides: CliConfigOverrides) -> Config {
+async fn load_config_or_exit(cli_config_overrides: CliConfigOverrides) -> Config {
     let cli_overrides = match cli_config_overrides.parse_overrides() {
         Ok(v) => v,
         Err(e) => {
@@ -142,7 +197,7 @@ fn load_config_or_exit(cli_config_overrides: CliConfigOverrides) -> Config {
     };
 
     let config_overrides = ConfigOverrides::default();
-    match Config::load_with_cli_overrides(cli_overrides, config_overrides) {
+    match Config::load_with_cli_overrides(cli_overrides, config_overrides).await {
         Ok(config) => config,
         Err(e) => {
             eprintln!("Error loading configuration: {e}");

codex-rs/cli/src/main.rs

@@ -7,6 +7,7 @@ use codex_chatgpt::apply_command::ApplyCommand;
 use codex_chatgpt::apply_command::run_apply_command;
 use codex_cli::LandlockCommand;
 use codex_cli::SeatbeltCommand;
+use codex_cli::login::read_api_key_from_stdin;
 use codex_cli::login::run_login_status;
 use codex_cli::login::run_login_with_api_key;
 use codex_cli::login::run_login_with_chatgpt;
@@ -18,6 +19,7 @@ use codex_exec::Cli as ExecCli;
 use codex_responses_api_proxy::Args as ResponsesApiProxyArgs;
 use codex_tui::AppExitInfo;
 use codex_tui::Cli as TuiCli;
+use codex_tui::UpdateAction;
 use owo_colors::OwoColorize;
 use std::path::PathBuf;
 use supports_color::Stream;
@@ -25,6 +27,8 @@ use supports_color::Stream;
 mod mcp_cmd;
 
 use crate::mcp_cmd::McpCli;
+use codex_core::config::Config;
+use codex_core::config::ConfigOverrides;
 
 /// Codex CLI
 ///
@@ -38,12 +42,16 @@ use crate::mcp_cmd::McpCli;
     // The executable is sometimes invoked via a platform‑specific name like
     // `codex-x86_64-unknown-linux-musl`, but the help output should always use
     // the generic `codex` command name that users run.
-    bin_name = "codex"
+    bin_name = "codex",
+    override_usage = "codex [OPTIONS] [PROMPT]\n       codex [OPTIONS] <COMMAND> [ARGS]"
 )]
 struct MultitoolCli {
     #[clap(flatten)]
     pub config_overrides: CliConfigOverrides,
 
+    #[clap(flatten)]
+    pub feature_toggles: FeatureToggles,
+
     #[clap(flatten)]
     interactive: TuiCli,
 
@@ -75,8 +83,9 @@ enum Subcommand {
     /// Generate shell completion scripts.
     Completion(CompletionCommand),
 
-    /// Internal debugging commands.
-    Debug(DebugArgs),
+    /// Run commands within a Codex-provided sandbox.
+    #[clap(visible_alias = "debug")]
+    Sandbox(SandboxArgs),
 
     /// Apply the latest diff produced by Codex agent as a `git apply` to your local working tree.
     #[clap(visible_alias = "a")]
@@ -95,6 +104,13 @@ enum Subcommand {
     /// Internal: run the responses API proxy.
     #[clap(hide = true)]
     ResponsesApiProxy(ResponsesApiProxyArgs),
+
+    /// Internal: relay stdio to a Unix domain socket.
+    #[clap(hide = true, name = "stdio-to-uds")]
+    StdioToUds(StdioToUdsCommand),
+
+    /// Inspect feature flags.
+    Features(FeaturesCli),
 }
 
 #[derive(Debug, Parser)]
@@ -120,18 +136,20 @@ struct ResumeCommand {
 }
 
 #[derive(Debug, Parser)]
-struct DebugArgs {
+struct SandboxArgs {
     #[command(subcommand)]
-    cmd: DebugCommand,
+    cmd: SandboxCommand,
 }
 
 #[derive(Debug, clap::Subcommand)]
-enum DebugCommand {
+enum SandboxCommand {
     /// Run a command under Seatbelt (macOS only).
-    Seatbelt(SeatbeltCommand),
+    #[clap(visible_alias = "seatbelt")]
+    Macos(SeatbeltCommand),
 
     /// Run a command under Landlock+seccomp (Linux only).
-    Landlock(LandlockCommand),
+    #[clap(visible_alias = "landlock")]
+    Linux(LandlockCommand),
 }
 
 #[derive(Debug, Parser)]
@@ -139,12 +157,21 @@ struct LoginCommand {
     #[clap(skip)]
     config_overrides: CliConfigOverrides,
 
-    #[arg(long = "api-key", value_name = "API_KEY")]
+    #[arg(
+        long = "with-api-key",
+        help = "Read the API key from stdin (e.g. `printenv OPENAI_API_KEY | codex login --with-api-key`)"
+    )]
+    with_api_key: bool,
+
+    #[arg(
+        long = "api-key",
+        value_name = "API_KEY",
+        help = "(deprecated) Previously accepted the API key directly; now exits with guidance to use --with-api-key",
+        hide = true
+    )]
     api_key: Option<String>,
 
-    /// EXPERIMENTAL: Use device code flow (not yet supported)
-    /// This feature is experimental and may changed in future releases.
-    #[arg(long = "experimental_use-device-code", hide = true)]
+    #[arg(long = "device-auth")]
     use_device_code: bool,
 
     /// EXPERIMENTAL: Use custom OAuth issuer base URL (advanced)
@@ -183,10 +210,18 @@ struct GenerateTsCommand {
     prettier: Option<PathBuf>,
 }
 
+#[derive(Debug, Parser)]
+struct StdioToUdsCommand {
+    /// Path to the Unix domain socket to connect to.
+    #[arg(value_name = "SOCKET_PATH")]
+    socket_path: PathBuf,
+}
+
 fn format_exit_messages(exit_info: AppExitInfo, color_enabled: bool) -> Vec<String> {
     let AppExitInfo {
         token_usage,
         conversation_id,
+        ..
     } = exit_info;
 
     if token_usage.is_zero() {
@@ -211,11 +246,79 @@ fn format_exit_messages(exit_info: AppExitInfo, color_enabled: bool) -> Vec<Stri
     lines
 }
 
-fn print_exit_messages(exit_info: AppExitInfo) {
+/// Handle the app exit and print the results. Optionally run the update action.
+fn handle_app_exit(exit_info: AppExitInfo) -> anyhow::Result<()> {
+    let update_action = exit_info.update_action;
     let color_enabled = supports_color::on(Stream::Stdout).is_some();
     for line in format_exit_messages(exit_info, color_enabled) {
         println!("{line}");
     }
+    if let Some(action) = update_action {
+        run_update_action(action)?;
+    }
+    Ok(())
+}
+
+/// Run the update action and print the result.
+fn run_update_action(action: UpdateAction) -> anyhow::Result<()> {
+    println!();
+    let (cmd, args) = action.command_args();
+    let cmd_str = action.command_str();
+    println!("Updating Codex via `{cmd_str}`...");
+    let status = std::process::Command::new(cmd).args(args).status()?;
+    if !status.success() {
+        anyhow::bail!("`{cmd_str}` failed with status {status}");
+    }
+    println!();
+    println!("🎉 Update ran successfully! Please restart Codex.");
+    Ok(())
+}
+
+#[derive(Debug, Default, Parser, Clone)]
+struct FeatureToggles {
+    /// Enable a feature (repeatable). Equivalent to `-c features.<name>=true`.
+    #[arg(long = "enable", value_name = "FEATURE", action = clap::ArgAction::Append, global = true)]
+    enable: Vec<String>,
+
+    /// Disable a feature (repeatable). Equivalent to `-c features.<name>=false`.
+    #[arg(long = "disable", value_name = "FEATURE", action = clap::ArgAction::Append, global = true)]
+    disable: Vec<String>,
+}
+
+impl FeatureToggles {
+    fn to_overrides(&self) -> Vec<String> {
+        let mut v = Vec::new();
+        for k in &self.enable {
+            v.push(format!("features.{k}=true"));
+        }
+        for k in &self.disable {
+            v.push(format!("features.{k}=false"));
+        }
+        v
+    }
+}
+
+#[derive(Debug, Parser)]
+struct FeaturesCli {
+    #[command(subcommand)]
+    sub: FeaturesSubcommand,
+}
+
+#[derive(Debug, Parser)]
+enum FeaturesSubcommand {
+    /// List known features with their stage and effective state.
+    List,
+}
+
+fn stage_str(stage: codex_core::features::Stage) -> &'static str {
+    use codex_core::features::Stage;
+    match stage {
+        Stage::Experimental => "experimental",
+        Stage::Beta => "beta",
+        Stage::Stable => "stable",
+        Stage::Deprecated => "deprecated",
+        Stage::Removed => "removed",
+    }
 }
 
 /// As early as possible in the process lifecycle, apply hardening measures. We
@@ -235,11 +338,17 @@ fn main() -> anyhow::Result<()> {
 
 async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()> {
     let MultitoolCli {
-        config_overrides: root_config_overrides,
+        config_overrides: mut root_config_overrides,
+        feature_toggles,
         mut interactive,
         subcommand,
     } = MultitoolCli::parse();
 
+    // Fold --enable/--disable into config overrides so they flow to all subcommands.
+    root_config_overrides
+        .raw_overrides
+        .extend(feature_toggles.to_overrides());
+
     match subcommand {
         None => {
             prepend_config_flags(
@@ -247,7 +356,7 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
                 root_config_overrides.clone(),
             );
             let exit_info = codex_tui::run_main(interactive, codex_linux_sandbox_exe).await?;
-            print_exit_messages(exit_info);
+            handle_app_exit(exit_info)?;
         }
         Some(Subcommand::Exec(mut exec_cli)) => {
             prepend_config_flags(
@@ -279,7 +388,8 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
                 last,
                 config_overrides,
             );
-            codex_tui::run_main(interactive, codex_linux_sandbox_exe).await?;
+            let exit_info = codex_tui::run_main(interactive, codex_linux_sandbox_exe).await?;
+            handle_app_exit(exit_info)?;
         }
         Some(Subcommand::Login(mut login_cli)) => {
             prepend_config_flags(
@@ -298,7 +408,13 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
                             login_cli.client_id,
                         )
                         .await;
-                    } else if let Some(api_key) = login_cli.api_key {
+                    } else if login_cli.api_key.is_some() {
+                        eprintln!(
+                            "The --api-key flag is no longer supported. Pipe the key instead, e.g. `printenv OPENAI_API_KEY | codex login --with-api-key`."
+                        );
+                        std::process::exit(1);
+                    } else if login_cli.with_api_key {
+                        let api_key = read_api_key_from_stdin();
                         run_login_with_api_key(login_cli.config_overrides, api_key).await;
                     } else {
                         run_login_with_chatgpt(login_cli.config_overrides).await;
@@ -323,8 +439,8 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
             );
             codex_cloud_tasks::run_main(cloud_cli, codex_linux_sandbox_exe).await?;
         }
-        Some(Subcommand::Debug(debug_args)) => match debug_args.cmd {
-            DebugCommand::Seatbelt(mut seatbelt_cli) => {
+        Some(Subcommand::Sandbox(sandbox_args)) => match sandbox_args.cmd {
+            SandboxCommand::Macos(mut seatbelt_cli) => {
                 prepend_config_flags(
                     &mut seatbelt_cli.config_overrides,
                     root_config_overrides.clone(),
@@ -335,7 +451,7 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
                 )
                 .await?;
             }
-            DebugCommand::Landlock(mut landlock_cli) => {
+            SandboxCommand::Linux(mut landlock_cli) => {
                 prepend_config_flags(
                     &mut landlock_cli.config_overrides,
                     root_config_overrides.clone(),
@@ -358,9 +474,38 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
             tokio::task::spawn_blocking(move || codex_responses_api_proxy::run_main(args))
                 .await??;
         }
+        Some(Subcommand::StdioToUds(cmd)) => {
+            let socket_path = cmd.socket_path;
+            tokio::task::spawn_blocking(move || codex_stdio_to_uds::run(socket_path.as_path()))
+                .await??;
+        }
         Some(Subcommand::GenerateTs(gen_cli)) => {
             codex_protocol_ts::generate_ts(&gen_cli.out_dir, gen_cli.prettier.as_deref())?;
         }
+        Some(Subcommand::Features(FeaturesCli { sub })) => match sub {
+            FeaturesSubcommand::List => {
+                // Respect root-level `-c` overrides plus top-level flags like `--profile`.
+                let cli_kv_overrides = root_config_overrides
+                    .parse_overrides()
+                    .map_err(|e| anyhow::anyhow!(e))?;
+
+                // Thread through relevant top-level flags (at minimum, `--profile`).
+                // Also honor `--search` since it maps to a feature toggle.
+                let overrides = ConfigOverrides {
+                    config_profile: interactive.config_profile.clone(),
+                    tools_web_search_request: interactive.web_search.then_some(true),
+                    ..Default::default()
+                };
+
+                let config = Config::load_with_cli_overrides(cli_kv_overrides, overrides).await?;
+                for def in codex_core::features::FEATURES.iter() {
+                    let name = def.key;
+                    let stage = stage_str(def.stage);
+                    let enabled = config.features.enabled(def.id);
+                    println!("{name}\t{stage}\t{enabled}");
+                }
+            }
+        },
     }
 
     Ok(())
@@ -435,6 +580,9 @@ fn merge_resume_cli_flags(interactive: &mut TuiCli, resume_cli: TuiCli) {
     if !resume_cli.images.is_empty() {
         interactive.images = resume_cli.images;
     }
+    if !resume_cli.add_dir.is_empty() {
+        interactive.add_dir.extend(resume_cli.add_dir);
+    }
     if let Some(prompt) = resume_cli.prompt {
         interactive.prompt = Some(prompt);
     }
@@ -454,6 +602,7 @@ fn print_completion(cmd: CompletionCommand) {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use assert_matches::assert_matches;
     use codex_core::protocol::TokenUsage;
     use codex_protocol::ConversationId;
 
@@ -463,6 +612,7 @@ mod tests {
             interactive,
             config_overrides: root_overrides,
             subcommand,
+            feature_toggles: _,
         } = cli;
 
         let Subcommand::Resume(ResumeCommand {
@@ -488,6 +638,7 @@ mod tests {
             conversation_id: conversation
                 .map(ConversationId::from_string)
                 .map(Result::unwrap),
+            update_action: None,
         }
     }
 
@@ -496,6 +647,7 @@ mod tests {
         let exit_info = AppExitInfo {
             token_usage: TokenUsage::default(),
             conversation_id: None,
+            update_action: None,
         };
         let lines = format_exit_messages(exit_info, false);
         assert!(lines.is_empty());
@@ -586,14 +738,14 @@ mod tests {
         assert_eq!(interactive.model.as_deref(), Some("gpt-5-test"));
         assert!(interactive.oss);
         assert_eq!(interactive.config_profile.as_deref(), Some("my-profile"));
-        assert!(matches!(
+        assert_matches!(
             interactive.sandbox_mode,
             Some(codex_common::SandboxModeCliArg::WorkspaceWrite)
-        ));
-        assert!(matches!(
+        );
+        assert_matches!(
             interactive.approval_policy,
             Some(codex_common::ApprovalModeCliArg::OnRequest)
-        ));
+        );
         assert!(interactive.full_auto);
         assert_eq!(
             interactive.cwd.as_deref(),

codex-rs/cli/src/mcp_cmd.rs

@@ -4,7 +4,9 @@ use anyhow::Context;
 use anyhow::Result;
 use anyhow::anyhow;
 use anyhow::bail;
+use clap::ArgGroup;
 use codex_common::CliConfigOverrides;
+use codex_common::format_env_display::format_env_display;
 use codex_core::config::Config;
 use codex_core::config::ConfigOverrides;
 use codex_core::config::find_codex_home;
@@ -12,6 +14,12 @@ use codex_core::config::load_global_mcp_servers;
 use codex_core::config::write_global_mcp_servers;
 use codex_core::config_types::McpServerConfig;
 use codex_core::config_types::McpServerTransportConfig;
+use codex_core::features::Feature;
+use codex_core::mcp::auth::compute_auth_statuses;
+use codex_core::protocol::McpAuthStatus;
+use codex_rmcp_client::delete_oauth_tokens;
+use codex_rmcp_client::perform_oauth_login;
+use codex_rmcp_client::supports_oauth_login;
 
 /// [experimental] Launch Codex as an MCP server or manage configured MCP servers.
 ///
@@ -43,6 +51,14 @@ pub enum McpSubcommand {
 
     /// [experimental] Remove a global MCP server entry.
     Remove(RemoveArgs),
+
+    /// [experimental] Authenticate with a configured MCP server via OAuth.
+    /// Requires experimental_use_rmcp_client = true in config.toml.
+    Login(LoginArgs),
+
+    /// [experimental] Remove stored OAuth credentials for a server.
+    /// Requires experimental_use_rmcp_client = true in config.toml.
+    Logout(LogoutArgs),
 }
 
 #[derive(Debug, clap::Parser)]
@@ -67,13 +83,61 @@ pub struct AddArgs {
     /// Name for the MCP server configuration.
     pub name: String,
 
-    /// Environment variables to set when launching the server.
-    #[arg(long, value_parser = parse_env_pair, value_name = "KEY=VALUE")]
-    pub env: Vec<(String, String)>,
+    #[command(flatten)]
+    pub transport_args: AddMcpTransportArgs,
+}
 
+#[derive(Debug, clap::Args)]
+#[command(
+    group(
+        ArgGroup::new("transport")
+            .args(["command", "url"])
+            .required(true)
+            .multiple(false)
+    )
+)]
+pub struct AddMcpTransportArgs {
+    #[command(flatten)]
+    pub stdio: Option<AddMcpStdioArgs>,
+
+    #[command(flatten)]
+    pub streamable_http: Option<AddMcpStreamableHttpArgs>,
+}
+
+#[derive(Debug, clap::Args)]
+pub struct AddMcpStdioArgs {
     /// Command to launch the MCP server.
-    #[arg(trailing_var_arg = true, num_args = 1..)]
+    /// Use --url for a streamable HTTP server.
+    #[arg(
+            trailing_var_arg = true,
+            num_args = 0..,
+        )]
     pub command: Vec<String>,
+
+    /// Environment variables to set when launching the server.
+    /// Only valid with stdio servers.
+    #[arg(
+        long,
+        value_parser = parse_env_pair,
+        value_name = "KEY=VALUE",
+    )]
+    pub env: Vec<(String, String)>,
+}
+
+#[derive(Debug, clap::Args)]
+pub struct AddMcpStreamableHttpArgs {
+    /// URL for a streamable HTTP MCP server.
+    #[arg(long)]
+    pub url: String,
+
+    /// Optional environment variable to read for a bearer token.
+    /// Only valid with streamable HTTP servers.
+    #[arg(
+        long = "bearer-token-env-var",
+        value_name = "ENV_VAR",
+        requires = "url"
+    )]
+    pub bearer_token_env_var: Option<String>,
 }
 
 #[derive(Debug, clap::Parser)]
@@ -82,6 +146,18 @@ pub struct RemoveArgs {
     pub name: String,
 }
 
+#[derive(Debug, clap::Parser)]
+pub struct LoginArgs {
+    /// Name of the MCP server to authenticate with oauth.
+    pub name: String,
+}
+
+#[derive(Debug, clap::Parser)]
+pub struct LogoutArgs {
+    /// Name of the MCP server to deauthenticate.
+    pub name: String,
+}
+
 impl McpCli {
     pub async fn run(self) -> Result<()> {
         let McpCli {
@@ -91,16 +167,22 @@ impl McpCli {
 
         match subcommand {
             McpSubcommand::List(args) => {
-                run_list(&config_overrides, args)?;
+                run_list(&config_overrides, args).await?;
             }
             McpSubcommand::Get(args) => {
-                run_get(&config_overrides, args)?;
+                run_get(&config_overrides, args).await?;
             }
             McpSubcommand::Add(args) => {
-                run_add(&config_overrides, args)?;
+                run_add(&config_overrides, args).await?;
             }
             McpSubcommand::Remove(args) => {
-                run_remove(&config_overrides, args)?;
+                run_remove(&config_overrides, args).await?;
+            }
+            McpSubcommand::Login(args) => {
+                run_login(&config_overrides, args).await?;
+            }
+            McpSubcommand::Logout(args) => {
+                run_logout(&config_overrides, args).await?;
             }
         }
 
@@ -108,40 +190,67 @@ impl McpCli {
     }
 }
 
-fn run_add(config_overrides: &CliConfigOverrides, add_args: AddArgs) -> Result<()> {
+async fn run_add(config_overrides: &CliConfigOverrides, add_args: AddArgs) -> Result<()> {
     // Validate any provided overrides even though they are not currently applied.
-    config_overrides.parse_overrides().map_err(|e| anyhow!(e))?;
+    let overrides = config_overrides.parse_overrides().map_err(|e| anyhow!(e))?;
+    let config = Config::load_with_cli_overrides(overrides, ConfigOverrides::default())
+        .await
+        .context("failed to load configuration")?;
 
-    let AddArgs { name, env, command } = add_args;
+    let AddArgs {
+        name,
+        transport_args,
+    } = add_args;
 
     validate_server_name(&name)?;
 
-    let mut command_parts = command.into_iter();
-    let command_bin = command_parts
-        .next()
-        .ok_or_else(|| anyhow!("command is required"))?;
-    let command_args: Vec<String> = command_parts.collect();
-
-    let env_map = if env.is_empty() {
-        None
-    } else {
-        let mut map = HashMap::new();
-        for (key, value) in env {
-            map.insert(key, value);
-        }
-        Some(map)
-    };
-
     let codex_home = find_codex_home().context("failed to resolve CODEX_HOME")?;
     let mut servers = load_global_mcp_servers(&codex_home)
+        .await
         .with_context(|| format!("failed to load MCP servers from {}", codex_home.display()))?;
 
-    let new_entry = McpServerConfig {
-        transport: McpServerTransportConfig::Stdio {
-            command: command_bin,
-            args: command_args,
-            env: env_map,
+    let transport = match transport_args {
+        AddMcpTransportArgs {
+            stdio: Some(stdio), ..
+        } => {
+            let mut command_parts = stdio.command.into_iter();
+            let command_bin = command_parts
+                .next()
+                .ok_or_else(|| anyhow!("command is required"))?;
+            let command_args: Vec<String> = command_parts.collect();
+
+            let env_map = if stdio.env.is_empty() {
+                None
+            } else {
+                Some(stdio.env.into_iter().collect::<HashMap<_, _>>())
+            };
+            McpServerTransportConfig::Stdio {
+                command: command_bin,
+                args: command_args,
+                env: env_map,
+                env_vars: Vec::new(),
+                cwd: None,
+            }
+        }
+        AddMcpTransportArgs {
+            streamable_http:
+                Some(AddMcpStreamableHttpArgs {
+                    url,
+                    bearer_token_env_var,
+                }),
+            ..
+        } => McpServerTransportConfig::StreamableHttp {
+            url,
+            bearer_token_env_var,
+            http_headers: None,
+            env_http_headers: None,
         },
+        AddMcpTransportArgs { .. } => bail!("exactly one of --command or --url must be provided"),
+    };
+
+    let new_entry = McpServerConfig {
+        transport: transport.clone(),
+        enabled: true,
         startup_timeout_sec: None,
         tool_timeout_sec: None,
     };
@@ -153,10 +262,30 @@ fn run_add(config_overrides: &CliConfigOverrides, add_args: AddArgs) -> Result<(
 
     println!("Added global MCP server '{name}'.");
 
+    if let McpServerTransportConfig::StreamableHttp {
+        url,
+        bearer_token_env_var: None,
+        http_headers,
+        env_http_headers,
+    } = transport
+        && matches!(supports_oauth_login(&url).await, Ok(true))
+    {
+        println!("Detected OAuth support. Starting OAuth flow…");
+        perform_oauth_login(
+            &name,
+            &url,
+            config.mcp_oauth_credentials_store_mode,
+            http_headers.clone(),
+            env_http_headers.clone(),
+        )
+        .await?;
+        println!("Successfully logged in.");
+    }
+
     Ok(())
 }
 
-fn run_remove(config_overrides: &CliConfigOverrides, remove_args: RemoveArgs) -> Result<()> {
+async fn run_remove(config_overrides: &CliConfigOverrides, remove_args: RemoveArgs) -> Result<()> {
     config_overrides.parse_overrides().map_err(|e| anyhow!(e))?;
 
     let RemoveArgs { name } = remove_args;
@@ -165,6 +294,7 @@ fn run_remove(config_overrides: &CliConfigOverrides, remove_args: RemoveArgs) ->
 
     let codex_home = find_codex_home().context("failed to resolve CODEX_HOME")?;
     let mut servers = load_global_mcp_servers(&codex_home)
+        .await
         .with_context(|| format!("failed to load MCP servers from {}", codex_home.display()))?;
 
     let removed = servers.remove(&name).is_some();
@@ -183,36 +313,129 @@ fn run_remove(config_overrides: &CliConfigOverrides, remove_args: RemoveArgs) ->
     Ok(())
 }
 
-fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) -> Result<()> {
+async fn run_login(config_overrides: &CliConfigOverrides, login_args: LoginArgs) -> Result<()> {
     let overrides = config_overrides.parse_overrides().map_err(|e| anyhow!(e))?;
     let config = Config::load_with_cli_overrides(overrides, ConfigOverrides::default())
+        .await
+        .context("failed to load configuration")?;
+
+    if !config.features.enabled(Feature::RmcpClient) {
+        bail!(
+            "OAuth login is only supported when experimental_use_rmcp_client is true in config.toml."
+        );
+    }
+
+    let LoginArgs { name } = login_args;
+
+    let Some(server) = config.mcp_servers.get(&name) else {
+        bail!("No MCP server named '{name}' found.");
+    };
+
+    let (url, http_headers, env_http_headers) = match &server.transport {
+        McpServerTransportConfig::StreamableHttp {
+            url,
+            http_headers,
+            env_http_headers,
+            ..
+        } => (url.clone(), http_headers.clone(), env_http_headers.clone()),
+        _ => bail!("OAuth login is only supported for streamable HTTP servers."),
+    };
+
+    perform_oauth_login(
+        &name,
+        &url,
+        config.mcp_oauth_credentials_store_mode,
+        http_headers,
+        env_http_headers,
+    )
+    .await?;
+    println!("Successfully logged in to MCP server '{name}'.");
+    Ok(())
+}
+
+async fn run_logout(config_overrides: &CliConfigOverrides, logout_args: LogoutArgs) -> Result<()> {
+    let overrides = config_overrides.parse_overrides().map_err(|e| anyhow!(e))?;
+    let config = Config::load_with_cli_overrides(overrides, ConfigOverrides::default())
+        .await
+        .context("failed to load configuration")?;
+
+    let LogoutArgs { name } = logout_args;
+
+    let server = config
+        .mcp_servers
+        .get(&name)
+        .ok_or_else(|| anyhow!("No MCP server named '{name}' found in configuration."))?;
+
+    let url = match &server.transport {
+        McpServerTransportConfig::StreamableHttp { url, .. } => url.clone(),
+        _ => bail!("OAuth logout is only supported for streamable_http transports."),
+    };
+
+    match delete_oauth_tokens(&name, &url, config.mcp_oauth_credentials_store_mode) {
+        Ok(true) => println!("Removed OAuth credentials for '{name}'."),
+        Ok(false) => println!("No OAuth credentials stored for '{name}'."),
+        Err(err) => return Err(anyhow!("failed to delete OAuth credentials: {err}")),
+    }
+
+    Ok(())
+}
+
+async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) -> Result<()> {
+    let overrides = config_overrides.parse_overrides().map_err(|e| anyhow!(e))?;
+    let config = Config::load_with_cli_overrides(overrides, ConfigOverrides::default())
+        .await
         .context("failed to load configuration")?;
 
     let mut entries: Vec<_> = config.mcp_servers.iter().collect();
     entries.sort_by(|(a, _), (b, _)| a.cmp(b));
+    let auth_statuses = compute_auth_statuses(
+        config.mcp_servers.iter(),
+        config.mcp_oauth_credentials_store_mode,
+    )
+    .await;
 
     if list_args.json {
         let json_entries: Vec<_> = entries
             .into_iter()
             .map(|(name, cfg)| {
+                let auth_status = auth_statuses
+                    .get(name.as_str())
+                    .copied()
+                    .unwrap_or(McpAuthStatus::Unsupported);
                 let transport = match &cfg.transport {
-                    McpServerTransportConfig::Stdio { command, args, env } => serde_json::json!({
+                    McpServerTransportConfig::Stdio {
+                        command,
+                        args,
+                        env,
+                        env_vars,
+                        cwd,
+                    } => serde_json::json!({
                         "type": "stdio",
                         "command": command,
                         "args": args,
                         "env": env,
+                        "env_vars": env_vars,
+                        "cwd": cwd,
                     }),
-                    McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
+                    McpServerTransportConfig::StreamableHttp {
+                        url,
+                        bearer_token_env_var,
+                        http_headers,
+                        env_http_headers,
+                    } => {
                         serde_json::json!({
                             "type": "streamable_http",
                             "url": url,
-                            "bearer_token": bearer_token,
+                            "bearer_token_env_var": bearer_token_env_var,
+                            "http_headers": http_headers,
+                            "env_http_headers": env_http_headers,
                         })
                     }
                 };
 
                 serde_json::json!({
                     "name": name,
+                    "enabled": cfg.enabled,
                     "transport": transport,
                     "startup_timeout_sec": cfg
                         .startup_timeout_sec
@@ -220,6 +443,7 @@ fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) -> Resul
                     "tool_timeout_sec": cfg
                         .tool_timeout_sec
                         .map(|timeout| timeout.as_secs_f64()),
+                    "auth_status": auth_status,
                 })
             })
             .collect();
@@ -233,45 +457,85 @@ fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) -> Resul
         return Ok(());
     }
 
-    let mut stdio_rows: Vec<[String; 4]> = Vec::new();
-    let mut http_rows: Vec<[String; 3]> = Vec::new();
+    let mut stdio_rows: Vec<[String; 7]> = Vec::new();
+    let mut http_rows: Vec<[String; 5]> = Vec::new();
 
     for (name, cfg) in entries {
         match &cfg.transport {
-            McpServerTransportConfig::Stdio { command, args, env } => {
+            McpServerTransportConfig::Stdio {
+                command,
+                args,
+                env,
+                env_vars,
+                cwd,
+            } => {
                 let args_display = if args.is_empty() {
                     "-".to_string()
                 } else {
                     args.join(" ")
                 };
-                let env_display = match env.as_ref() {
-                    None => "-".to_string(),
-                    Some(map) if map.is_empty() => "-".to_string(),
-                    Some(map) => {
-                        let mut pairs: Vec<_> = map.iter().collect();
-                        pairs.sort_by(|(a, _), (b, _)| a.cmp(b));
-                        pairs
-                            .into_iter()
-                            .map(|(k, v)| format!("{k}={v}"))
-                            .collect::<Vec<_>>()
-                            .join(", ")
-                    }
-                };
-                stdio_rows.push([name.clone(), command.clone(), args_display, env_display]);
-            }
-            McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
-                let has_bearer = if bearer_token.is_some() {
-                    "True"
+                let env_display = format_env_display(env.as_ref(), env_vars);
+                let cwd_display = cwd
+                    .as_ref()
+                    .map(|path| path.display().to_string())
+                    .filter(|value| !value.is_empty())
+                    .unwrap_or_else(|| "-".to_string());
+                let status = if cfg.enabled {
+                    "enabled".to_string()
                 } else {
-                    "False"
+                    "disabled".to_string()
                 };
-                http_rows.push([name.clone(), url.clone(), has_bearer.into()]);
+                let auth_status = auth_statuses
+                    .get(name.as_str())
+                    .copied()
+                    .unwrap_or(McpAuthStatus::Unsupported)
+                    .to_string();
+                stdio_rows.push([
+                    name.clone(),
+                    command.clone(),
+                    args_display,
+                    env_display,
+                    cwd_display,
+                    status,
+                    auth_status,
+                ]);
+            }
+            McpServerTransportConfig::StreamableHttp {
+                url,
+                bearer_token_env_var,
+                ..
+            } => {
+                let status = if cfg.enabled {
+                    "enabled".to_string()
+                } else {
+                    "disabled".to_string()
+                };
+                let auth_status = auth_statuses
+                    .get(name.as_str())
+                    .copied()
+                    .unwrap_or(McpAuthStatus::Unsupported)
+                    .to_string();
+                http_rows.push([
+                    name.clone(),
+                    url.clone(),
+                    bearer_token_env_var.clone().unwrap_or("-".to_string()),
+                    status,
+                    auth_status,
+                ]);
             }
         }
     }
 
     if !stdio_rows.is_empty() {
-        let mut widths = ["Name".len(), "Command".len(), "Args".len(), "Env".len()];
+        let mut widths = [
+            "Name".len(),
+            "Command".len(),
+            "Args".len(),
+            "Env".len(),
+            "Cwd".len(),
+            "Status".len(),
+            "Auth".len(),
+        ];
         for row in &stdio_rows {
             for (i, cell) in row.iter().enumerate() {
                 widths[i] = widths[i].max(cell.len());
@@ -279,28 +543,40 @@ fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) -> Resul
         }
 
         println!(
-            "{:<name_w$}  {:<cmd_w$}  {:<args_w$}  {:<env_w$}",
-            "Name",
-            "Command",
-            "Args",
-            "Env",
+            "{name:<name_w$}  {command:<cmd_w$}  {args:<args_w$}  {env:<env_w$}  {cwd:<cwd_w$}  {status:<status_w$}  {auth:<auth_w$}",
+            name = "Name",
+            command = "Command",
+            args = "Args",
+            env = "Env",
+            cwd = "Cwd",
+            status = "Status",
+            auth = "Auth",
             name_w = widths[0],
             cmd_w = widths[1],
             args_w = widths[2],
             env_w = widths[3],
+            cwd_w = widths[4],
+            status_w = widths[5],
+            auth_w = widths[6],
         );
 
         for row in &stdio_rows {
             println!(
-                "{:<name_w$}  {:<cmd_w$}  {:<args_w$}  {:<env_w$}",
-                row[0],
-                row[1],
-                row[2],
-                row[3],
+                "{name:<name_w$}  {command:<cmd_w$}  {args:<args_w$}  {env:<env_w$}  {cwd:<cwd_w$}  {status:<status_w$}  {auth:<auth_w$}",
+                name = row[0].as_str(),
+                command = row[1].as_str(),
+                args = row[2].as_str(),
+                env = row[3].as_str(),
+                cwd = row[4].as_str(),
+                status = row[5].as_str(),
+                auth = row[6].as_str(),
                 name_w = widths[0],
                 cmd_w = widths[1],
                 args_w = widths[2],
                 env_w = widths[3],
+                cwd_w = widths[4],
+                status_w = widths[5],
+                auth_w = widths[6],
             );
         }
     }
@@ -310,7 +586,13 @@ fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) -> Resul
     }
 
     if !http_rows.is_empty() {
-        let mut widths = ["Name".len(), "Url".len(), "Has Bearer Token".len()];
+        let mut widths = [
+            "Name".len(),
+            "Url".len(),
+            "Bearer Token Env Var".len(),
+            "Status".len(),
+            "Auth".len(),
+        ];
         for row in &http_rows {
             for (i, cell) in row.iter().enumerate() {
                 widths[i] = widths[i].max(cell.len());
@@ -318,24 +600,32 @@ fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) -> Resul
         }
 
         println!(
-            "{:<name_w$}  {:<url_w$}  {:<token_w$}",
-            "Name",
-            "Url",
-            "Has Bearer Token",
+            "{name:<name_w$}  {url:<url_w$}  {token:<token_w$}  {status:<status_w$}  {auth:<auth_w$}",
+            name = "Name",
+            url = "Url",
+            token = "Bearer Token Env Var",
+            status = "Status",
+            auth = "Auth",
             name_w = widths[0],
             url_w = widths[1],
             token_w = widths[2],
+            status_w = widths[3],
+            auth_w = widths[4],
         );
 
         for row in &http_rows {
             println!(
-                "{:<name_w$}  {:<url_w$}  {:<token_w$}",
-                row[0],
-                row[1],
-                row[2],
+                "{name:<name_w$}  {url:<url_w$}  {token:<token_w$}  {status:<status_w$}  {auth:<auth_w$}",
+                name = row[0].as_str(),
+                url = row[1].as_str(),
+                token = row[2].as_str(),
+                status = row[3].as_str(),
+                auth = row[4].as_str(),
                 name_w = widths[0],
                 url_w = widths[1],
                 token_w = widths[2],
+                status_w = widths[3],
+                auth_w = widths[4],
             );
         }
     }
@@ -343,9 +633,10 @@ fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) -> Resul
     Ok(())
 }
 
-fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Result<()> {
+async fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Result<()> {
     let overrides = config_overrides.parse_overrides().map_err(|e| anyhow!(e))?;
     let config = Config::load_with_cli_overrides(overrides, ConfigOverrides::default())
+        .await
         .context("failed to load configuration")?;
 
     let Some(server) = config.mcp_servers.get(&get_args.name) else {
@@ -354,20 +645,36 @@ fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Result<(
 
     if get_args.json {
         let transport = match &server.transport {
-            McpServerTransportConfig::Stdio { command, args, env } => serde_json::json!({
+            McpServerTransportConfig::Stdio {
+                command,
+                args,
+                env,
+                env_vars,
+                cwd,
+            } => serde_json::json!({
                 "type": "stdio",
                 "command": command,
                 "args": args,
                 "env": env,
+                "env_vars": env_vars,
+                "cwd": cwd,
             }),
-            McpServerTransportConfig::StreamableHttp { url, bearer_token } => serde_json::json!({
+            McpServerTransportConfig::StreamableHttp {
+                url,
+                bearer_token_env_var,
+                http_headers,
+                env_http_headers,
+            } => serde_json::json!({
                 "type": "streamable_http",
                 "url": url,
-                "bearer_token": bearer_token,
+                "bearer_token_env_var": bearer_token_env_var,
+                "http_headers": http_headers,
+                "env_http_headers": env_http_headers,
             }),
         };
         let output = serde_json::to_string_pretty(&serde_json::json!({
             "name": get_args.name,
+            "enabled": server.enabled,
             "transport": transport,
             "startup_timeout_sec": server
                 .startup_timeout_sec
@@ -381,8 +688,15 @@ fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Result<(
     }
 
     println!("{}", get_args.name);
+    println!("  enabled: {}", server.enabled);
     match &server.transport {
-        McpServerTransportConfig::Stdio { command, args, env } => {
+        McpServerTransportConfig::Stdio {
+            command,
+            args,
+            env,
+            env_vars,
+            cwd,
+        } => {
             println!("  transport: stdio");
             println!("  command: {command}");
             let args_display = if args.is_empty() {
@@ -391,10 +705,27 @@ fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Result<(
                 args.join(" ")
             };
             println!("  args: {args_display}");
-            let env_display = match env.as_ref() {
-                None => "-".to_string(),
-                Some(map) if map.is_empty() => "-".to_string(),
-                Some(map) => {
+            let cwd_display = cwd
+                .as_ref()
+                .map(|path| path.display().to_string())
+                .filter(|value| !value.is_empty())
+                .unwrap_or_else(|| "-".to_string());
+            println!("  cwd: {cwd_display}");
+            let env_display = format_env_display(env.as_ref(), env_vars);
+            println!("  env: {env_display}");
+        }
+        McpServerTransportConfig::StreamableHttp {
+            url,
+            bearer_token_env_var,
+            http_headers,
+            env_http_headers,
+        } => {
+            println!("  transport: streamable_http");
+            println!("  url: {url}");
+            let env_var = bearer_token_env_var.as_deref().unwrap_or("-");
+            println!("  bearer_token_env_var: {env_var}");
+            let headers_display = match http_headers {
+                Some(map) if !map.is_empty() => {
                     let mut pairs: Vec<_> = map.iter().collect();
                     pairs.sort_by(|(a, _), (b, _)| a.cmp(b));
                     pairs
@@ -403,14 +734,22 @@ fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Result<(
                         .collect::<Vec<_>>()
                         .join(", ")
                 }
+                _ => "-".to_string(),
             };
-            println!("  env: {env_display}");
-        }
-        McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
-            println!("  transport: streamable_http");
-            println!("  url: {url}");
-            let bearer = bearer_token.as_deref().unwrap_or("-");
-            println!("  bearer_token: {bearer}");
+            println!("  http_headers: {headers_display}");
+            let env_headers_display = match env_http_headers {
+                Some(map) if !map.is_empty() => {
+                    let mut pairs: Vec<_> = map.iter().collect();
+                    pairs.sort_by(|(a, _), (b, _)| a.cmp(b));
+                    pairs
+                        .into_iter()
+                        .map(|(k, v)| format!("{k}={v}"))
+                        .collect::<Vec<_>>()
+                        .join(", ")
+                }
+                _ => "-".to_string(),
+            };
+            println!("  env_http_headers: {env_headers_display}");
         }
     }
     if let Some(timeout) = server.startup_timeout_sec {

codex-rs/cli/tests/mcp_add_remove.rs

@@ -13,8 +13,8 @@ fn codex_command(codex_home: &Path) -> Result<assert_cmd::Command> {
     Ok(cmd)
 }
 
-#[test]
-fn add_and_remove_server_updates_global_config() -> Result<()> {
+#[tokio::test]
+async fn add_and_remove_server_updates_global_config() -> Result<()> {
     let codex_home = TempDir::new()?;
 
     let mut add_cmd = codex_command(codex_home.path())?;
@@ -24,17 +24,26 @@ fn add_and_remove_server_updates_global_config() -> Result<()> {
         .success()
         .stdout(contains("Added global MCP server 'docs'."));
 
-    let servers = load_global_mcp_servers(codex_home.path())?;
+    let servers = load_global_mcp_servers(codex_home.path()).await?;
     assert_eq!(servers.len(), 1);
     let docs = servers.get("docs").expect("server should exist");
     match &docs.transport {
-        McpServerTransportConfig::Stdio { command, args, env } => {
+        McpServerTransportConfig::Stdio {
+            command,
+            args,
+            env,
+            env_vars,
+            cwd,
+        } => {
             assert_eq!(command, "echo");
             assert_eq!(args, &vec!["hello".to_string()]);
             assert!(env.is_none());
+            assert!(env_vars.is_empty());
+            assert!(cwd.is_none());
         }
         other => panic!("unexpected transport: {other:?}"),
     }
+    assert!(docs.enabled);
 
     let mut remove_cmd = codex_command(codex_home.path())?;
     remove_cmd
@@ -43,7 +52,7 @@ fn add_and_remove_server_updates_global_config() -> Result<()> {
         .success()
         .stdout(contains("Removed global MCP server 'docs'."));
 
-    let servers = load_global_mcp_servers(codex_home.path())?;
+    let servers = load_global_mcp_servers(codex_home.path()).await?;
     assert!(servers.is_empty());
 
     let mut remove_again_cmd = codex_command(codex_home.path())?;
@@ -53,14 +62,14 @@ fn add_and_remove_server_updates_global_config() -> Result<()> {
         .success()
         .stdout(contains("No MCP server named 'docs' found."));
 
-    let servers = load_global_mcp_servers(codex_home.path())?;
+    let servers = load_global_mcp_servers(codex_home.path()).await?;
     assert!(servers.is_empty());
 
     Ok(())
 }
 
-#[test]
-fn add_with_env_preserves_key_order_and_values() -> Result<()> {
+#[tokio::test]
+async fn add_with_env_preserves_key_order_and_values() -> Result<()> {
     let codex_home = TempDir::new()?;
 
     let mut add_cmd = codex_command(codex_home.path())?;
@@ -80,7 +89,7 @@ fn add_with_env_preserves_key_order_and_values() -> Result<()> {
         .assert()
         .success();
 
-    let servers = load_global_mcp_servers(codex_home.path())?;
+    let servers = load_global_mcp_servers(codex_home.path()).await?;
     let envy = servers.get("envy").expect("server should exist");
     let env = match &envy.transport {
         McpServerTransportConfig::Stdio { env: Some(env), .. } => env,
@@ -90,6 +99,130 @@ fn add_with_env_preserves_key_order_and_values() -> Result<()> {
     assert_eq!(env.len(), 2);
     assert_eq!(env.get("FOO"), Some(&"bar".to_string()));
     assert_eq!(env.get("ALPHA"), Some(&"beta".to_string()));
+    assert!(envy.enabled);
+
+    Ok(())
+}
+
+#[tokio::test]
+async fn add_streamable_http_without_manual_token() -> Result<()> {
+    let codex_home = TempDir::new()?;
+
+    let mut add_cmd = codex_command(codex_home.path())?;
+    add_cmd
+        .args(["mcp", "add", "github", "--url", "https://example.com/mcp"])
+        .assert()
+        .success();
+
+    let servers = load_global_mcp_servers(codex_home.path()).await?;
+    let github = servers.get("github").expect("github server should exist");
+    match &github.transport {
+        McpServerTransportConfig::StreamableHttp {
+            url,
+            bearer_token_env_var,
+            http_headers,
+            env_http_headers,
+        } => {
+            assert_eq!(url, "https://example.com/mcp");
+            assert!(bearer_token_env_var.is_none());
+            assert!(http_headers.is_none());
+            assert!(env_http_headers.is_none());
+        }
+        other => panic!("unexpected transport: {other:?}"),
+    }
+    assert!(github.enabled);
+
+    assert!(!codex_home.path().join(".credentials.json").exists());
+    assert!(!codex_home.path().join(".env").exists());
+
+    Ok(())
+}
+
+#[tokio::test]
+async fn add_streamable_http_with_custom_env_var() -> Result<()> {
+    let codex_home = TempDir::new()?;
+
+    let mut add_cmd = codex_command(codex_home.path())?;
+    add_cmd
+        .args([
+            "mcp",
+            "add",
+            "issues",
+            "--url",
+            "https://example.com/issues",
+            "--bearer-token-env-var",
+            "GITHUB_TOKEN",
+        ])
+        .assert()
+        .success();
+
+    let servers = load_global_mcp_servers(codex_home.path()).await?;
+    let issues = servers.get("issues").expect("issues server should exist");
+    match &issues.transport {
+        McpServerTransportConfig::StreamableHttp {
+            url,
+            bearer_token_env_var,
+            http_headers,
+            env_http_headers,
+        } => {
+            assert_eq!(url, "https://example.com/issues");
+            assert_eq!(bearer_token_env_var.as_deref(), Some("GITHUB_TOKEN"));
+            assert!(http_headers.is_none());
+            assert!(env_http_headers.is_none());
+        }
+        other => panic!("unexpected transport: {other:?}"),
+    }
+    assert!(issues.enabled);
+    Ok(())
+}
+
+#[tokio::test]
+async fn add_streamable_http_rejects_removed_flag() -> Result<()> {
+    let codex_home = TempDir::new()?;
+
+    let mut add_cmd = codex_command(codex_home.path())?;
+    add_cmd
+        .args([
+            "mcp",
+            "add",
+            "github",
+            "--url",
+            "https://example.com/mcp",
+            "--with-bearer-token",
+        ])
+        .assert()
+        .failure()
+        .stderr(contains("--with-bearer-token"));
+
+    let servers = load_global_mcp_servers(codex_home.path()).await?;
+    assert!(servers.is_empty());
+
+    Ok(())
+}
+
+#[tokio::test]
+async fn add_cant_add_command_and_url() -> Result<()> {
+    let codex_home = TempDir::new()?;
+
+    let mut add_cmd = codex_command(codex_home.path())?;
+    add_cmd
+        .args([
+            "mcp",
+            "add",
+            "github",
+            "--url",
+            "https://example.com/mcp",
+            "--command",
+            "--",
+            "echo",
+            "hello",
+        ])
+        .assert()
+        .failure()
+        .stderr(contains("unexpected argument '--command' found"));
+
+    let servers = load_global_mcp_servers(codex_home.path()).await?;
+    assert!(servers.is_empty());
 
     Ok(())
 }

codex-rs/cli/tests/mcp_list.rs

@@ -1,6 +1,10 @@
 use std::path::Path;
 
 use anyhow::Result;
+use codex_core::config::load_global_mcp_servers;
+use codex_core::config::write_global_mcp_servers;
+use codex_core::config_types::McpServerTransportConfig;
+use predicates::prelude::PredicateBooleanExt;
 use predicates::str::contains;
 use pretty_assertions::assert_eq;
 use serde_json::Value as JsonValue;
@@ -26,8 +30,8 @@ fn list_shows_empty_state() -> Result<()> {
     Ok(())
 }
 
-#[test]
-fn list_and_get_render_expected_output() -> Result<()> {
+#[tokio::test]
+async fn list_and_get_render_expected_output() -> Result<()> {
     let codex_home = TempDir::new()?;
 
     let mut add = codex_command(codex_home.path())?;
@@ -45,6 +49,18 @@ fn list_and_get_render_expected_output() -> Result<()> {
     .assert()
     .success();
 
+    let mut servers = load_global_mcp_servers(codex_home.path()).await?;
+    let docs_entry = servers
+        .get_mut("docs")
+        .expect("docs server should exist after add");
+    match &mut docs_entry.transport {
+        McpServerTransportConfig::Stdio { env_vars, .. } => {
+            *env_vars = vec!["APP_TOKEN".to_string(), "WORKSPACE_ID".to_string()];
+        }
+        other => panic!("unexpected transport: {other:?}"),
+    }
+    write_global_mcp_servers(codex_home.path(), &servers)?;
+
     let mut list_cmd = codex_command(codex_home.path())?;
     let list_output = list_cmd.args(["mcp", "list"]).output()?;
     assert!(list_output.status.success());
@@ -53,6 +69,12 @@ fn list_and_get_render_expected_output() -> Result<()> {
     assert!(stdout.contains("docs"));
     assert!(stdout.contains("docs-server"));
     assert!(stdout.contains("TOKEN=secret"));
+    assert!(stdout.contains("APP_TOKEN=$APP_TOKEN"));
+    assert!(stdout.contains("WORKSPACE_ID=$WORKSPACE_ID"));
+    assert!(stdout.contains("Status"));
+    assert!(stdout.contains("Auth"));
+    assert!(stdout.contains("enabled"));
+    assert!(stdout.contains("Unsupported"));
 
     let mut list_json_cmd = codex_command(codex_home.path())?;
     let json_output = list_json_cmd.args(["mcp", "list", "--json"]).output()?;
@@ -64,6 +86,7 @@ fn list_and_get_render_expected_output() -> Result<()> {
         json!([
           {
             "name": "docs",
+            "enabled": true,
             "transport": {
               "type": "stdio",
               "command": "docs-server",
@@ -73,10 +96,16 @@ fn list_and_get_render_expected_output() -> Result<()> {
               ],
               "env": {
                 "TOKEN": "secret"
-              }
+              },
+              "env_vars": [
+                "APP_TOKEN",
+                "WORKSPACE_ID"
+              ],
+              "cwd": null
             },
             "startup_timeout_sec": null,
-            "tool_timeout_sec": null
+            "tool_timeout_sec": null,
+            "auth_status": "unsupported"
           }
         ]
         )
@@ -91,6 +120,9 @@ fn list_and_get_render_expected_output() -> Result<()> {
     assert!(stdout.contains("command: docs-server"));
     assert!(stdout.contains("args: --port 4000"));
     assert!(stdout.contains("env: TOKEN=secret"));
+    assert!(stdout.contains("APP_TOKEN=$APP_TOKEN"));
+    assert!(stdout.contains("WORKSPACE_ID=$WORKSPACE_ID"));
+    assert!(stdout.contains("enabled: true"));
     assert!(stdout.contains("remove: codex mcp remove docs"));
 
     let mut get_json_cmd = codex_command(codex_home.path())?;
@@ -98,7 +130,7 @@ fn list_and_get_render_expected_output() -> Result<()> {
         .args(["mcp", "get", "docs", "--json"])
         .assert()
         .success()
-        .stdout(contains("\"name\": \"docs\""));
+        .stdout(contains("\"name\": \"docs\"").and(contains("\"enabled\": true")));
 
     Ok(())
 }

codex-rs/clippy.toml

@@ -7,3 +7,7 @@ disallowed-methods = [
     { path = "ratatui::style::Stylize::black", reason = "Avoid hardcoding black; prefer default fg or dim/bold. Exception: Disable this rule if rendering over a hardcoded ANSI background." },
     { path = "ratatui::style::Stylize::yellow", reason = "Avoid yellow; prefer other colors in `tui/styles.md`." },
 ]
+
+# Increase the size threshold for result_large_err to accommodate
+# richer error variants.
+large-error-threshold = 256

codex-rs/cloud-tasks/Cargo.toml

@@ -1,7 +1,7 @@
 [package]
+edition = "2024"
 name = "codex-cloud-tasks"
 version = { workspace = true }
-edition = "2024"
 
 [lib]
 name = "codex_cloud_tasks"
@@ -11,26 +11,28 @@ path = "src/lib.rs"
 workspace = true
 
 [dependencies]
-anyhow = "1"
-clap = { version = "4", features = ["derive"] }
+anyhow = { workspace = true }
+base64 = { workspace = true }
+chrono = { workspace = true, features = ["serde"] }
+clap = { workspace = true, features = ["derive"] }
+codex-cloud-tasks-client = { path = "../cloud-tasks-client", features = [
+    "mock",
+    "online",
+] }
 codex-common = { path = "../common", features = ["cli"] }
-tokio = { version = "1", features = ["macros", "rt-multi-thread"] }
-tracing = { version = "0.1.41", features = ["log"] }
-tracing-subscriber = { version = "0.3.19", features = ["env-filter"] }
-codex-cloud-tasks-client = { path = "../cloud-tasks-client", features = ["mock", "online"] }
-ratatui = { version = "0.29.0" }
-crossterm = { version = "0.28.1", features = ["event-stream"] }
-tokio-stream = "0.1.17"
-chrono = { version = "0.4", features = ["serde"] }
-codex-login = { path = "../login" }
 codex-core = { path = "../core" }
-throbber-widgets-tui = "0.8.0"
-base64 = "0.22"
-serde_json = "1"
-reqwest = { version = "0.12", features = ["json"] }
-serde = { version = "1", features = ["derive"] }
-unicode-width = "0.1"
+codex-login = { path = "../login" }
 codex-tui = { path = "../tui" }
+crossterm = { workspace = true, features = ["event-stream"] }
+ratatui = { workspace = true }
+reqwest = { workspace = true, features = ["json"] }
+serde = { workspace = true, features = ["derive"] }
+serde_json = { workspace = true }
+tokio = { workspace = true, features = ["macros", "rt-multi-thread"] }
+tokio-stream = { workspace = true }
+tracing = { workspace = true, features = ["log"] }
+tracing-subscriber = { workspace = true, features = ["env-filter"] }
+unicode-width = { workspace = true }
 
 [dev-dependencies]
-async-trait = "0.1"
+async-trait = { workspace = true }

codex-rs/cloud-tasks/src/app.rs

@@ -1,4 +1,5 @@
 use std::time::Duration;
+use std::time::Instant;
 
 // Environment filter data models for the TUI
 #[derive(Clone, Debug, Default)]
@@ -42,15 +43,13 @@ use crate::scrollable_diff::ScrollableDiff;
 use codex_cloud_tasks_client::CloudBackend;
 use codex_cloud_tasks_client::TaskId;
 use codex_cloud_tasks_client::TaskSummary;
-use throbber_widgets_tui::ThrobberState;
-
 #[derive(Default)]
 pub struct App {
     pub tasks: Vec<TaskSummary>,
     pub selected: usize,
     pub status: String,
     pub diff_overlay: Option<DiffOverlay>,
-    pub throbber: ThrobberState,
+    pub spinner_start: Option<Instant>,
     pub refresh_inflight: bool,
     pub details_inflight: bool,
     // Environment filter state
@@ -82,7 +81,7 @@ impl App {
             selected: 0,
             status: "Press r to refresh".to_string(),
             diff_overlay: None,
-            throbber: ThrobberState::default(),
+            spinner_start: None,
             refresh_inflight: false,
             details_inflight: false,
             env_filter: None,

codex-rs/cloud-tasks/src/cli.rs

@@ -1,3 +1,4 @@
+use clap::Args;
 use clap::Parser;
 use codex_common::CliConfigOverrides;
 
@@ -6,4 +7,43 @@ use codex_common::CliConfigOverrides;
 pub struct Cli {
     #[clap(skip)]
     pub config_overrides: CliConfigOverrides,
+
+    #[command(subcommand)]
+    pub command: Option<Command>,
+}
+
+#[derive(Debug, clap::Subcommand)]
+pub enum Command {
+    /// Submit a new Codex Cloud task without launching the TUI.
+    Exec(ExecCommand),
+}
+
+#[derive(Debug, Args)]
+pub struct ExecCommand {
+    /// Task prompt to run in Codex Cloud.
+    #[arg(value_name = "QUERY")]
+    pub query: Option<String>,
+
+    /// Target environment identifier (see `codex cloud` to browse).
+    #[arg(long = "env", value_name = "ENV_ID")]
+    pub environment: String,
+
+    /// Number of assistant attempts (best-of-N).
+    #[arg(
+        long = "attempts",
+        default_value_t = 1usize,
+        value_parser = parse_attempts
+    )]
+    pub attempts: usize,
+}
+
+fn parse_attempts(input: &str) -> Result<usize, String> {
+    let value: usize = input
+        .parse()
+        .map_err(|_| "attempts must be an integer between 1 and 4".to_string())?;
+    if (1..=4).contains(&value) {
+        Ok(value)
+    } else {
+        Err("attempts must be between 1 and 4".to_string())
+    }
 }

codex-rs/cloud-tasks/src/lib.rs

@@ -7,7 +7,9 @@ mod ui;
 pub mod util;
 pub use cli::Cli;
 
+use anyhow::anyhow;
 use std::io::IsTerminal;
+use std::io::Read;
 use std::path::PathBuf;
 use std::sync::Arc;
 use std::time::Duration;
@@ -23,6 +25,175 @@ struct ApplyJob {
     diff_override: Option<String>,
 }
 
+struct BackendContext {
+    backend: Arc<dyn codex_cloud_tasks_client::CloudBackend>,
+    base_url: String,
+}
+
+async fn init_backend(user_agent_suffix: &str) -> anyhow::Result<BackendContext> {
+    let use_mock = matches!(
+        std::env::var("CODEX_CLOUD_TASKS_MODE").ok().as_deref(),
+        Some("mock") | Some("MOCK")
+    );
+    let base_url = std::env::var("CODEX_CLOUD_TASKS_BASE_URL")
+        .unwrap_or_else(|_| "https://chatgpt.com/backend-api".to_string());
+
+    set_user_agent_suffix(user_agent_suffix);
+
+    if use_mock {
+        return Ok(BackendContext {
+            backend: Arc::new(codex_cloud_tasks_client::MockClient),
+            base_url,
+        });
+    }
+
+    let ua = codex_core::default_client::get_codex_user_agent();
+    let mut http = codex_cloud_tasks_client::HttpClient::new(base_url.clone())?.with_user_agent(ua);
+    let style = if base_url.contains("/backend-api") {
+        "wham"
+    } else {
+        "codex-api"
+    };
+    append_error_log(format!("startup: base_url={base_url} path_style={style}"));
+
+    let auth = match codex_core::config::find_codex_home()
+        .ok()
+        .map(|home| codex_login::AuthManager::new(home, false))
+        .and_then(|am| am.auth())
+    {
+        Some(auth) => auth,
+        None => {
+            eprintln!(
+                "Not signed in. Please run 'codex login' to sign in with ChatGPT, then re-run 'codex cloud'."
+            );
+            std::process::exit(1);
+        }
+    };
+
+    if let Some(acc) = auth.get_account_id() {
+        append_error_log(format!("auth: mode=ChatGPT account_id={acc}"));
+    }
+
+    let token = match auth.get_token().await {
+        Ok(t) if !t.is_empty() => t,
+        _ => {
+            eprintln!(
+                "Not signed in. Please run 'codex login' to sign in with ChatGPT, then re-run 'codex cloud'."
+            );
+            std::process::exit(1);
+        }
+    };
+
+    http = http.with_bearer_token(token.clone());
+    if let Some(acc) = auth
+        .get_account_id()
+        .or_else(|| util::extract_chatgpt_account_id(&token))
+    {
+        append_error_log(format!("auth: set ChatGPT-Account-Id header: {acc}"));
+        http = http.with_chatgpt_account_id(acc);
+    }
+
+    Ok(BackendContext {
+        backend: Arc::new(http),
+        base_url,
+    })
+}
+
+async fn run_exec_command(args: crate::cli::ExecCommand) -> anyhow::Result<()> {
+    let crate::cli::ExecCommand {
+        query,
+        environment,
+        attempts,
+    } = args;
+    let ctx = init_backend("codex_cloud_tasks_exec").await?;
+    let prompt = resolve_query_input(query)?;
+    let env_id = resolve_environment_id(&ctx, &environment).await?;
+    let created = codex_cloud_tasks_client::CloudBackend::create_task(
+        &*ctx.backend,
+        &env_id,
+        &prompt,
+        "main",
+        false,
+        attempts,
+    )
+    .await?;
+    let url = util::task_url(&ctx.base_url, &created.id.0);
+    println!("{url}");
+    Ok(())
+}
+
+async fn resolve_environment_id(ctx: &BackendContext, requested: &str) -> anyhow::Result<String> {
+    let trimmed = requested.trim();
+    if trimmed.is_empty() {
+        return Err(anyhow!("environment id must not be empty"));
+    }
+    let normalized = util::normalize_base_url(&ctx.base_url);
+    let headers = util::build_chatgpt_headers().await;
+    let environments = crate::env_detect::list_environments(&normalized, &headers).await?;
+    if environments.is_empty() {
+        return Err(anyhow!(
+            "no cloud environments are available for this workspace"
+        ));
+    }
+
+    if let Some(row) = environments.iter().find(|row| row.id == trimmed) {
+        return Ok(row.id.clone());
+    }
+
+    let label_matches = environments
+        .iter()
+        .filter(|row| {
+            row.label
+                .as_deref()
+                .map(|label| label.eq_ignore_ascii_case(trimmed))
+                .unwrap_or(false)
+        })
+        .collect::<Vec<_>>();
+    match label_matches.as_slice() {
+        [] => Err(anyhow!(
+            "environment '{trimmed}' not found; run `codex cloud` to list available environments"
+        )),
+        [single] => Ok(single.id.clone()),
+        [first, rest @ ..] => {
+            let first_id = &first.id;
+            if rest.iter().all(|row| row.id == *first_id) {
+                Ok(first_id.clone())
+            } else {
+                Err(anyhow!(
+                    "environment label '{trimmed}' is ambiguous; run `codex cloud` to pick the desired environment id"
+                ))
+            }
+        }
+    }
+}
+
+fn resolve_query_input(query_arg: Option<String>) -> anyhow::Result<String> {
+    match query_arg {
+        Some(q) if q != "-" => Ok(q),
+        maybe_dash => {
+            let force_stdin = matches!(maybe_dash.as_deref(), Some("-"));
+            if std::io::stdin().is_terminal() && !force_stdin {
+                return Err(anyhow!(
+                    "no query provided. Pass one as an argument or pipe it via stdin."
+                ));
+            }
+            if !force_stdin {
+                eprintln!("Reading query from stdin...");
+            }
+            let mut buffer = String::new();
+            std::io::stdin()
+                .read_to_string(&mut buffer)
+                .map_err(|e| anyhow!("failed to read query from stdin: {e}"))?;
+            if buffer.trim().is_empty() {
+                return Err(anyhow!(
+                    "no query provided via stdin (received empty input)."
+                ));
+            }
+            Ok(buffer)
+        }
+    }
+}
+
 fn level_from_status(status: codex_cloud_tasks_client::ApplyStatus) -> app::ApplyResultLevel {
     match status {
         codex_cloud_tasks_client::ApplyStatus::Success => app::ApplyResultLevel::Success,
@@ -148,7 +319,14 @@ fn spawn_apply(
 // (no standalone patch summarizer needed – UI displays raw diffs)
 
 /// Entry point for the `codex cloud` subcommand.
-pub async fn run_main(_cli: Cli, _codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()> {
+pub async fn run_main(cli: Cli, _codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()> {
+    if let Some(command) = cli.command {
+        return match command {
+            crate::cli::Command::Exec(args) => run_exec_command(args).await,
+        };
+    }
+    let Cli { .. } = cli;
+
     // Very minimal logging setup; mirrors other crates' pattern.
     let default_level = "error";
     let _ = tracing_subscriber::fmt()
@@ -162,72 +340,8 @@ pub async fn run_main(_cli: Cli, _codex_linux_sandbox_exe: Option<PathBuf>) -> a
         .try_init();
 
     info!("Launching Cloud Tasks list UI");
-    set_user_agent_suffix("codex_cloud_tasks_tui");
-
-    // Default to online unless explicitly configured to use mock.
-    let use_mock = matches!(
-        std::env::var("CODEX_CLOUD_TASKS_MODE").ok().as_deref(),
-        Some("mock") | Some("MOCK")
-    );
-
-    let backend: Arc<dyn codex_cloud_tasks_client::CloudBackend> = if use_mock {
-        Arc::new(codex_cloud_tasks_client::MockClient)
-    } else {
-        // Build an HTTP client against the configured (or default) base URL.
-        let base_url = std::env::var("CODEX_CLOUD_TASKS_BASE_URL")
-            .unwrap_or_else(|_| "https://chatgpt.com/backend-api".to_string());
-        let ua = codex_core::default_client::get_codex_user_agent();
-        let mut http =
-            codex_cloud_tasks_client::HttpClient::new(base_url.clone())?.with_user_agent(ua);
-        // Log which base URL and path style we're going to use.
-        let style = if base_url.contains("/backend-api") {
-            "wham"
-        } else {
-            "codex-api"
-        };
-        append_error_log(format!("startup: base_url={base_url} path_style={style}"));
-
-        // Require ChatGPT login (SWIC). Exit with a clear message if missing.
-        let _token = match codex_core::config::find_codex_home()
-            .ok()
-            .map(codex_login::AuthManager::new)
-            .and_then(|am| am.auth())
-        {
-            Some(auth) => {
-                // Log account context for debugging workspace selection.
-                if let Some(acc) = auth.get_account_id() {
-                    append_error_log(format!("auth: mode=ChatGPT account_id={acc}"));
-                }
-                match auth.get_token().await {
-                    Ok(t) if !t.is_empty() => {
-                        // Attach token and ChatGPT-Account-Id header if available
-                        http = http.with_bearer_token(t.clone());
-                        if let Some(acc) = auth
-                            .get_account_id()
-                            .or_else(|| util::extract_chatgpt_account_id(&t))
-                        {
-                            append_error_log(format!("auth: set ChatGPT-Account-Id header: {acc}"));
-                            http = http.with_chatgpt_account_id(acc);
-                        }
-                        t
-                    }
-                    _ => {
-                        eprintln!(
-                            "Not signed in. Please run 'codex login' to sign in with ChatGPT, then re-run 'codex cloud'."
-                        );
-                        std::process::exit(1);
-                    }
-                }
-            }
-            None => {
-                eprintln!(
-                    "Not signed in. Please run 'codex login' to sign in with ChatGPT, then re-run 'codex cloud'."
-                );
-                std::process::exit(1);
-            }
-        };
-        Arc::new(http)
-    };
+    let BackendContext { backend, .. } = init_backend("codex_cloud_tasks_tui").await?;
+    let backend = backend;
 
     // Terminal setup
     use crossterm::ExecutableCommand;
@@ -400,16 +514,20 @@ pub async fn run_main(_cli: Cli, _codex_linux_sandbox_exe: Option<PathBuf>) -> a
                         let _ = frame_tx.send(Instant::now() + codex_tui::ComposerInput::recommended_flush_delay());
                     }
                 }
-                // Advance throbber only while loading.
+                // Keep spinner pulsing only while loading.
                 if app.refresh_inflight
                     || app.details_inflight
                     || app.env_loading
                     || app.apply_preflight_inflight
                     || app.apply_inflight
                 {
-                    app.throbber.calc_next();
+                    if app.spinner_start.is_none() {
+                        app.spinner_start = Some(Instant::now());
+                    }
                     needs_redraw = true;
-                    let _ = frame_tx.send(Instant::now() + Duration::from_millis(100));
+                    let _ = frame_tx.send(Instant::now() + Duration::from_millis(600));
+                } else {
+                    app.spinner_start = None;
                 }
                 render_if_needed(&mut terminal, &mut app, &mut needs_redraw)?;
             }

codex-rs/cloud-tasks/src/ui.rs

@@ -16,6 +16,7 @@ use ratatui::widgets::ListState;
 use ratatui::widgets::Padding;
 use ratatui::widgets::Paragraph;
 use std::sync::OnceLock;
+use std::time::Instant;
 
 use crate::app::App;
 use crate::app::AttemptView;
@@ -229,7 +230,7 @@ fn draw_list(frame: &mut Frame, area: Rect, app: &mut App) {
 
     // In-box spinner during initial/refresh loads
     if app.refresh_inflight {
-        draw_centered_spinner(frame, inner, &mut app.throbber, "Loading tasks…");
+        draw_centered_spinner(frame, inner, &mut app.spinner_start, "Loading tasks…");
     }
 }
 
@@ -291,7 +292,7 @@ fn draw_footer(frame: &mut Frame, area: Rect, app: &mut App) {
         || app.apply_preflight_inflight
         || app.apply_inflight
     {
-        draw_inline_spinner(frame, top[1], &mut app.throbber, "Loading…");
+        draw_inline_spinner(frame, top[1], &mut app.spinner_start, "Loading…");
     } else {
         frame.render_widget(Clear, top[1]);
     }
@@ -449,7 +450,12 @@ fn draw_diff_overlay(frame: &mut Frame, area: Rect, app: &mut App) {
         .map(|o| o.sd.wrapped_lines().is_empty())
         .unwrap_or(true);
     if app.details_inflight && raw_empty {
-        draw_centered_spinner(frame, content_area, &mut app.throbber, "Loading details…");
+        draw_centered_spinner(
+            frame,
+            content_area,
+            &mut app.spinner_start,
+            "Loading details…",
+        );
     } else {
         let scroll = app
             .diff_overlay
@@ -494,11 +500,11 @@ pub fn draw_apply_modal(frame: &mut Frame, area: Rect, app: &mut App) {
         frame.render_widget(header, rows[0]);
         // Body: spinner while preflight/apply runs; otherwise show result message and path lists
         if app.apply_preflight_inflight {
-            draw_centered_spinner(frame, rows[1], &mut app.throbber, "Checking…");
+            draw_centered_spinner(frame, rows[1], &mut app.spinner_start, "Checking…");
         } else if app.apply_inflight {
-            draw_centered_spinner(frame, rows[1], &mut app.throbber, "Applying…");
+            draw_centered_spinner(frame, rows[1], &mut app.spinner_start, "Applying…");
         } else if m.result_message.is_none() {
-            draw_centered_spinner(frame, rows[1], &mut app.throbber, "Loading…");
+            draw_centered_spinner(frame, rows[1], &mut app.spinner_start, "Loading…");
         } else if let Some(msg) = &m.result_message {
             let mut body_lines: Vec<Line> = Vec::new();
             let first = match m.result_level {
@@ -859,29 +865,29 @@ fn format_relative_time(ts: chrono::DateTime<Utc>) -> String {
 fn draw_inline_spinner(
     frame: &mut Frame,
     area: Rect,
-    state: &mut throbber_widgets_tui::ThrobberState,
+    spinner_start: &mut Option<Instant>,
     label: &str,
 ) {
-    use ratatui::style::Style;
-    use throbber_widgets_tui::BRAILLE_EIGHT;
-    use throbber_widgets_tui::Throbber;
-    use throbber_widgets_tui::WhichUse;
-    let w = Throbber::default()
-        .label(label)
-        .style(Style::default().cyan())
-        .throbber_style(Style::default().magenta().bold())
-        .throbber_set(BRAILLE_EIGHT)
-        .use_type(WhichUse::Spin);
-    frame.render_stateful_widget(w, area, state);
+    use ratatui::widgets::Paragraph;
+    let start = spinner_start.get_or_insert_with(Instant::now);
+    let blink_on = (start.elapsed().as_millis() / 600).is_multiple_of(2);
+    let dot = if blink_on {
+        "• ".into()
+    } else {
+        "◦ ".dim()
+    };
+    let label = label.cyan();
+    let line = Line::from(vec![dot, label]);
+    frame.render_widget(Paragraph::new(line), area);
 }
 
 fn draw_centered_spinner(
     frame: &mut Frame,
     area: Rect,
-    state: &mut throbber_widgets_tui::ThrobberState,
+    spinner_start: &mut Option<Instant>,
     label: &str,
 ) {
-    // Center a 1xN throbber within the given rect
+    // Center a 1xN spinner within the given rect
     let rows = Layout::default()
         .direction(Direction::Vertical)
         .constraints([
@@ -898,7 +904,7 @@ fn draw_centered_spinner(
             Constraint::Percentage(50),
         ])
         .split(rows[1]);
-    draw_inline_spinner(frame, cols[1], state, label);
+    draw_inline_spinner(frame, cols[1], spinner_start, label);
 }
 
 // Styling helpers for diff rendering live inline where used.
@@ -918,7 +924,12 @@ pub fn draw_env_modal(frame: &mut Frame, area: Rect, app: &mut App) {
     let content = overlay_content(inner);
 
     if app.env_loading {
-        draw_centered_spinner(frame, content, &mut app.throbber, "Loading environments…");
+        draw_centered_spinner(
+            frame,
+            content,
+            &mut app.spinner_start,
+            "Loading environments…",
+        );
         return;
     }
 

codex-rs/cloud-tasks/src/util.rs

@@ -70,7 +70,7 @@ pub async fn build_chatgpt_headers() -> HeaderMap {
         HeaderValue::from_str(&ua).unwrap_or(HeaderValue::from_static("codex-cli")),
     );
     if let Ok(home) = codex_core::config::find_codex_home() {
-        let am = codex_login::AuthManager::new(home);
+        let am = codex_login::AuthManager::new(home, false);
         if let Some(auth) = am.auth()
             && let Ok(tok) = auth.get_token().await
             && !tok.is_empty()
@@ -91,3 +91,18 @@ pub async fn build_chatgpt_headers() -> HeaderMap {
     }
     headers
 }
+
+/// Construct a browser-friendly task URL for the given backend base URL.
+pub fn task_url(base_url: &str, task_id: &str) -> String {
+    let normalized = normalize_base_url(base_url);
+    if let Some(root) = normalized.strip_suffix("/backend-api") {
+        return format!("{root}/codex/tasks/{task_id}");
+    }
+    if let Some(root) = normalized.strip_suffix("/api/codex") {
+        return format!("{root}/codex/tasks/{task_id}");
+    }
+    if normalized.ends_with("/codex") {
+        return format!("{normalized}/tasks/{task_id}");
+    }
+    format!("{normalized}/codex/tasks/{task_id}")
+}

codex-rs/common/src/format_env_display.rs

@@ -0,0 +1,66 @@
+use std::collections::HashMap;
+
+pub fn format_env_display(env: Option<&HashMap<String, String>>, env_vars: &[String]) -> String {
+    let mut parts: Vec<String> = Vec::new();
+
+    if let Some(map) = env {
+        let mut pairs: Vec<_> = map.iter().collect();
+        pairs.sort_by(|(a, _), (b, _)| a.cmp(b));
+        parts.extend(
+            pairs
+                .into_iter()
+                .map(|(key, value)| format!("{key}={value}")),
+        );
+    }
+
+    if !env_vars.is_empty() {
+        parts.extend(env_vars.iter().map(|var| format!("{var}=${var}")));
+    }
+
+    if parts.is_empty() {
+        "-".to_string()
+    } else {
+        parts.join(", ")
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn returns_dash_when_empty() {
+        assert_eq!(format_env_display(None, &[]), "-");
+
+        let empty_map = HashMap::new();
+        assert_eq!(format_env_display(Some(&empty_map), &[]), "-");
+    }
+
+    #[test]
+    fn formats_sorted_env_pairs() {
+        let mut env = HashMap::new();
+        env.insert("B".to_string(), "two".to_string());
+        env.insert("A".to_string(), "one".to_string());
+
+        assert_eq!(format_env_display(Some(&env), &[]), "A=one, B=two");
+    }
+
+    #[test]
+    fn formats_env_vars_with_dollar_prefix() {
+        let vars = vec!["TOKEN".to_string(), "PATH".to_string()];
+
+        assert_eq!(format_env_display(None, &vars), "TOKEN=$TOKEN, PATH=$PATH");
+    }
+
+    #[test]
+    fn combines_env_pairs_and_vars() {
+        let mut env = HashMap::new();
+        env.insert("HOME".to_string(), "/tmp".to_string());
+        let vars = vec!["TOKEN".to_string()];
+
+        assert_eq!(
+            format_env_display(Some(&env), &vars),
+            "HOME=/tmp, TOKEN=$TOKEN"
+        );
+    }
+}

codex-rs/common/src/lib.rs

@@ -13,6 +13,9 @@ mod sandbox_mode_cli_arg;
 #[cfg(feature = "cli")]
 pub use sandbox_mode_cli_arg::SandboxModeCliArg;
 
+#[cfg(feature = "cli")]
+pub mod format_env_display;
+
 #[cfg(any(feature = "cli", test))]
 mod config_override;
 

codex-rs/common/src/model_presets.rs

@@ -20,49 +20,49 @@ const PRESETS: &[ModelPreset] = &[
     ModelPreset {
         id: "gpt-5-codex-low",
         label: "gpt-5-codex low",
-        description: "",
+        description: "Fastest responses with limited reasoning",
         model: "gpt-5-codex",
         effort: Some(ReasoningEffort::Low),
     },
     ModelPreset {
         id: "gpt-5-codex-medium",
         label: "gpt-5-codex medium",
-        description: "",
+        description: "Dynamically adjusts reasoning based on the task",
         model: "gpt-5-codex",
         effort: Some(ReasoningEffort::Medium),
     },
     ModelPreset {
         id: "gpt-5-codex-high",
         label: "gpt-5-codex high",
-        description: "",
+        description: "Maximizes reasoning depth for complex or ambiguous problems",
         model: "gpt-5-codex",
         effort: Some(ReasoningEffort::High),
     },
     ModelPreset {
         id: "gpt-5-minimal",
         label: "gpt-5 minimal",
-        description: "— fastest responses with limited reasoning; ideal for coding, instructions, or lightweight tasks",
+        description: "Fastest responses with little reasoning",
         model: "gpt-5",
         effort: Some(ReasoningEffort::Minimal),
     },
     ModelPreset {
         id: "gpt-5-low",
         label: "gpt-5 low",
-        description: "— balances speed with some reasoning; useful for straightforward queries and short explanations",
+        description: "Balances speed with some reasoning; useful for straightforward queries and short explanations",
         model: "gpt-5",
         effort: Some(ReasoningEffort::Low),
     },
     ModelPreset {
         id: "gpt-5-medium",
         label: "gpt-5 medium",
-        description: "— default setting; provides a solid balance of reasoning depth and latency for general-purpose tasks",
+        description: "Provides a solid balance of reasoning depth and latency for general-purpose tasks",
         model: "gpt-5",
         effort: Some(ReasoningEffort::Medium),
     },
     ModelPreset {
         id: "gpt-5-high",
         label: "gpt-5 high",
-        description: "— maximizes reasoning depth for complex or ambiguous problems",
+        description: "Maximizes reasoning depth for complex or ambiguous problems",
         model: "gpt-5",
         effort: Some(ReasoningEffort::High),
     },

codex-rs/config.md

@@ -3,4 +3,4 @@
 This file has moved. Please see the latest configuration documentation here:
 
 - Full config docs: [docs/config.md](../docs/config.md)
-- MCP servers section: [docs/config.md#mcp_servers](../docs/config.md#mcp_servers) 
+- MCP servers section: [docs/config.md#connecting-to-mcp-servers](../docs/config.md#connecting-to-mcp-servers)

codex-rs/core/Cargo.toml

@@ -19,13 +19,16 @@ async-trait = { workspace = true }
 base64 = { workspace = true }
 bytes = { workspace = true }
 chrono = { workspace = true, features = ["serde"] }
+codex-app-server-protocol = { workspace = true }
 codex-apply-patch = { workspace = true }
 codex-file-search = { workspace = true }
 codex-mcp-client = { workspace = true }
-codex-rmcp-client = { workspace = true }
-codex-protocol = { workspace = true }
-codex-app-server-protocol = { workspace = true }
 codex-otel = { workspace = true, features = ["otel"] }
+codex-protocol = { workspace = true }
+codex-rmcp-client = { workspace = true }
+codex-async-utils = { workspace = true }
+codex-utils-string = { workspace = true }
+codex-utils-pty = { workspace = true }
 dirs = { workspace = true }
 dunce = { workspace = true }
 env-flags = { workspace = true }
@@ -35,7 +38,6 @@ indexmap = { workspace = true }
 libc = { workspace = true }
 mcp-types = { workspace = true }
 os_info = { workspace = true }
-portable-pty = { workspace = true }
 rand = { workspace = true }
 regex-lite = { workspace = true }
 reqwest = { workspace = true, features = ["json", "stream"] }
@@ -46,6 +48,7 @@ shlex = { workspace = true }
 similar = { workspace = true }
 strum_macros = { workspace = true }
 tempfile = { workspace = true }
+test-log = { workspace = true }
 thiserror = { workspace = true }
 time = { workspace = true, features = [
     "formatting",
@@ -60,7 +63,7 @@ tokio = { workspace = true, features = [
     "rt-multi-thread",
     "signal",
 ] }
-tokio-util = { workspace = true }
+tokio-util = { workspace = true, features = ["rt"] }
 toml = { workspace = true }
 toml_edit = { workspace = true }
 tracing = { workspace = true, features = ["log"] }
@@ -75,6 +78,9 @@ wildmatch = { workspace = true }
 landlock = { workspace = true }
 seccompiler = { workspace = true }
 
+[target.'cfg(target_os = "macos")'.dependencies]
+core-foundation = "0.9"
+
 # Build OpenSSL from source for musl builds.
 [target.x86_64-unknown-linux-musl.dependencies]
 openssl-sys = { workspace = true, features = ["vendored"] }
@@ -85,16 +91,18 @@ openssl-sys = { workspace = true, features = ["vendored"] }
 
 [dev-dependencies]
 assert_cmd = { workspace = true }
+assert_matches = { workspace = true }
 core_test_support = { workspace = true }
 escargot = { workspace = true }
 maplit = { workspace = true }
 predicates = { workspace = true }
 pretty_assertions = { workspace = true }
+serial_test = { workspace = true }
 tempfile = { workspace = true }
 tokio-test = { workspace = true }
+tracing-test = { workspace = true, features = ["no-env-filter"] }
 walkdir = { workspace = true }
 wiremock = { workspace = true }
-tracing-test = {  workspace = true, features = ["no-env-filter"] }
 
 [package.metadata.cargo-shear]
 ignored = ["openssl-sys"]

codex-rs/core/README.md

@@ -4,7 +4,7 @@ This crate implements the business logic for Codex. It is designed to be used by
 
 ## Dependencies
 
-Note that `codex-core` makes some assumptions about certain helper utilities being available in the environment. Currently, this
+Note that `codex-core` makes some assumptions about certain helper utilities being available in the environment. Currently, this support matrix is:
 
 ### macOS
 
@@ -12,7 +12,7 @@ Expects `/usr/bin/sandbox-exec` to be present.
 
 ### Linux
 
-Expects the binary containing `codex-core` to run the equivalent of `codex debug landlock` when `arg0` is `codex-linux-sandbox`. See the `codex-arg0` crate for details.
+Expects the binary containing `codex-core` to run the equivalent of `codex sandbox linux` (legacy alias: `codex debug landlock`) when `arg0` is `codex-linux-sandbox`. See the `codex-arg0` crate for details.
 
 ### All Platforms
 

codex-rs/core/gpt_5_codex_prompt.md

@@ -10,12 +10,14 @@ You are Codex, based on GPT-5. You are running as a coding agent in the Codex CL
 
 - Default to ASCII when editing or creating files. Only introduce non-ASCII or other Unicode characters when there is a clear justification and the file already uses them.
 - Add succinct code comments that explain what is going on if code is not self-explanatory. You should not add comments like "Assigns the value to the variable", but a brief comment might be useful ahead of a complex code block that the user would otherwise have to spend time parsing out. Usage of these comments should be rare.
+- Try to use apply_patch for single file edits, but it is fine to explore other options to make the edit if it does not work well. Do not use apply_patch for changes that are auto-generated (i.e. generating package.json or running a lint or format command like gofmt) or when scripting is more efficient (such as search and replacing a string across a codebase).
 - You may be in a dirty git worktree.
     * NEVER revert existing changes you did not make unless explicitly requested, since these changes were made by the user.
     * If asked to make a commit or code edits and there are unrelated changes to your work or changes that you didn't make in those files, don't revert those changes.
     * If the changes are in files you've touched recently, you should read carefully and understand how you can work with the changes rather than reverting them.
     * If the changes are in unrelated files, just ignore them and don't revert them.
 - While you are working, you might notice unexpected changes that you didn't make. If this happens, STOP IMMEDIATELY and ask the user how they would like to proceed.
+- **NEVER** use destructive commands like `git reset --hard` or `git checkout --` unless specifically requested or approved by the user.
 
 ## Plan tool
 
@@ -89,7 +91,7 @@ You are producing plain text that will later be styled by the CLI. Follow these
 - Headers: optional; short Title Case (1-3 words) wrapped in **…**; no blank line before the first bullet; add only if they truly help.
 - Bullets: use - ; merge related points; keep to one line when possible; 4–6 per list ordered by importance; keep phrasing consistent.
 - Monospace: backticks for commands/paths/env vars/code ids and inline examples; use for literal keyword bullets; never combine with **.
-- Code samples or multi-line snippets should be wrapped in fenced code blocks; add a language hint whenever obvious.
+- Code samples or multi-line snippets should be wrapped in fenced code blocks; include an info string as often as possible.
 - Structure: group related bullets; order sections general → specific → supporting; for subsections, start with a bolded keyword bullet, then items; match complexity to the task.
 - Tone: collaborative, concise, factual; present tense, active voice; self‑contained; no "above/below"; parallel wording.
 - Don'ts: no nested bullets/hierarchies; no ANSI codes; don't cram unrelated keywords; keep keyword lists short—wrap/reformat if long; avoid naming formatting styles in answers.

codex-rs/core/src/auth.rs

@@ -2,6 +2,8 @@ use chrono::DateTime;
 use chrono::Utc;
 use serde::Deserialize;
 use serde::Serialize;
+#[cfg(test)]
+use serial_test::serial;
 use std::env;
 use std::fs::File;
 use std::fs::OpenOptions;
@@ -16,7 +18,9 @@ use std::sync::Mutex;
 use std::time::Duration;
 
 use codex_app_server_protocol::AuthMode;
+use codex_protocol::config_types::ForcedLoginMethod;
 
+use crate::config::Config;
 use crate::token_data::PlanType;
 use crate::token_data::TokenData;
 use crate::token_data::parse_id_token;
@@ -73,7 +77,7 @@ impl CodexAuth {
 
     /// Loads the available auth information from the auth.json.
     pub fn from_codex_home(codex_home: &Path) -> std::io::Result<Option<CodexAuth>> {
-        load_auth(codex_home)
+        load_auth(codex_home, false)
     }
 
     pub async fn get_token_data(&self) -> Result<TokenData, std::io::Error> {
@@ -135,6 +139,10 @@ impl CodexAuth {
         self.get_current_token_data().and_then(|t| t.account_id)
     }
 
+    pub fn get_account_email(&self) -> Option<String> {
+        self.get_current_token_data().and_then(|t| t.id_token.email)
+    }
+
     pub(crate) fn get_plan_type(&self) -> Option<PlanType> {
         self.get_current_token_data()
             .and_then(|t| t.id_token.chatgpt_plan_type)
@@ -188,6 +196,7 @@ impl CodexAuth {
 }
 
 pub const OPENAI_API_KEY_ENV_VAR: &str = "OPENAI_API_KEY";
+pub const CODEX_API_KEY_ENV_VAR: &str = "CODEX_API_KEY";
 
 pub fn read_openai_api_key_from_env() -> Option<String> {
     env::var(OPENAI_API_KEY_ENV_VAR)
@@ -196,6 +205,13 @@ pub fn read_openai_api_key_from_env() -> Option<String> {
         .filter(|value| !value.is_empty())
 }
 
+pub fn read_codex_api_key_from_env() -> Option<String> {
+    env::var(CODEX_API_KEY_ENV_VAR)
+        .ok()
+        .map(|value| value.trim().to_string())
+        .filter(|value| !value.is_empty())
+}
+
 pub fn get_auth_file(codex_home: &Path) -> PathBuf {
     codex_home.join("auth.json")
 }
@@ -221,14 +237,92 @@ pub fn login_with_api_key(codex_home: &Path, api_key: &str) -> std::io::Result<(
     write_auth_json(&get_auth_file(codex_home), &auth_dot_json)
 }
 
-fn load_auth(codex_home: &Path) -> std::io::Result<Option<CodexAuth>> {
+pub async fn enforce_login_restrictions(config: &Config) -> std::io::Result<()> {
+    let Some(auth) = load_auth(&config.codex_home, true)? else {
+        return Ok(());
+    };
+
+    if let Some(required_method) = config.forced_login_method {
+        let method_violation = match (required_method, auth.mode) {
+            (ForcedLoginMethod::Api, AuthMode::ApiKey) => None,
+            (ForcedLoginMethod::Chatgpt, AuthMode::ChatGPT) => None,
+            (ForcedLoginMethod::Api, AuthMode::ChatGPT) => Some(
+                "API key login is required, but ChatGPT is currently being used. Logging out."
+                    .to_string(),
+            ),
+            (ForcedLoginMethod::Chatgpt, AuthMode::ApiKey) => Some(
+                "ChatGPT login is required, but an API key is currently being used. Logging out."
+                    .to_string(),
+            ),
+        };
+
+        if let Some(message) = method_violation {
+            return logout_with_message(&config.codex_home, message);
+        }
+    }
+
+    if let Some(expected_account_id) = config.forced_chatgpt_workspace_id.as_deref() {
+        if auth.mode != AuthMode::ChatGPT {
+            return Ok(());
+        }
+
+        let token_data = match auth.get_token_data().await {
+            Ok(data) => data,
+            Err(err) => {
+                return logout_with_message(
+                    &config.codex_home,
+                    format!(
+                        "Failed to load ChatGPT credentials while enforcing workspace restrictions: {err}. Logging out."
+                    ),
+                );
+            }
+        };
+
+        // workspace is the external identifier for account id.
+        let chatgpt_account_id = token_data.id_token.chatgpt_account_id.as_deref();
+        if chatgpt_account_id != Some(expected_account_id) {
+            let message = match chatgpt_account_id {
+                Some(actual) => format!(
+                    "Login is restricted to workspace {expected_account_id}, but current credentials belong to {actual}. Logging out."
+                ),
+                None => format!(
+                    "Login is restricted to workspace {expected_account_id}, but current credentials lack a workspace identifier. Logging out."
+                ),
+            };
+            return logout_with_message(&config.codex_home, message);
+        }
+    }
+
+    Ok(())
+}
+
+fn logout_with_message(codex_home: &Path, message: String) -> std::io::Result<()> {
+    match logout(codex_home) {
+        Ok(_) => Err(std::io::Error::other(message)),
+        Err(err) => Err(std::io::Error::other(format!(
+            "{message}. Failed to remove auth.json: {err}"
+        ))),
+    }
+}
+
+fn load_auth(
+    codex_home: &Path,
+    enable_codex_api_key_env: bool,
+) -> std::io::Result<Option<CodexAuth>> {
+    if enable_codex_api_key_env && let Some(api_key) = read_codex_api_key_from_env() {
+        let client = crate::default_client::create_client();
+        return Ok(Some(CodexAuth::from_api_key_with_client(
+            api_key.as_str(),
+            client,
+        )));
+    }
+
     let auth_file = get_auth_file(codex_home);
     let client = crate::default_client::create_client();
     let auth_dot_json = match try_read_auth_json(&auth_file) {
         Ok(auth) => auth,
-        Err(e) => {
-            return Err(e);
-        }
+        Err(err) if err.kind() == std::io::ErrorKind::NotFound => return Ok(None),
+        Err(err) => return Err(err),
     };
 
     let AuthDotJson {
@@ -380,17 +474,19 @@ struct CachedAuth {
 #[cfg(test)]
 mod tests {
     use super::*;
+    use crate::config::Config;
+    use crate::config::ConfigOverrides;
+    use crate::config::ConfigToml;
     use crate::token_data::IdTokenInfo;
     use crate::token_data::KnownPlan;
     use crate::token_data::PlanType;
     use base64::Engine;
+    use codex_protocol::config_types::ForcedLoginMethod;
     use pretty_assertions::assert_eq;
     use serde::Serialize;
     use serde_json::json;
     use tempfile::tempdir;
 
-    const LAST_REFRESH: &str = "2025-08-06T20:41:36.232376Z";
-
     #[tokio::test]
     async fn roundtrip_auth_dot_json() {
         let codex_home = tempdir().unwrap();
@@ -398,6 +494,7 @@ mod tests {
             AuthFileParams {
                 openai_api_key: None,
                 chatgpt_plan_type: "pro".to_string(),
+                chatgpt_account_id: None,
             },
             codex_home.path(),
         )
@@ -437,6 +534,13 @@ mod tests {
         assert!(auth.tokens.is_none(), "tokens should be cleared");
     }
 
+    #[test]
+    fn missing_auth_json_returns_none() {
+        let dir = tempdir().unwrap();
+        let auth = CodexAuth::from_codex_home(dir.path()).expect("call should succeed");
+        assert_eq!(auth, None);
+    }
+
     #[tokio::test]
     async fn pro_account_with_no_api_key_uses_chatgpt_auth() {
         let codex_home = tempdir().unwrap();
@@ -444,6 +548,7 @@ mod tests {
             AuthFileParams {
                 openai_api_key: None,
                 chatgpt_plan_type: "pro".to_string(),
+                chatgpt_account_id: None,
             },
             codex_home.path(),
         )
@@ -455,12 +560,16 @@ mod tests {
             auth_dot_json,
             auth_file: _,
             ..
-        } = super::load_auth(codex_home.path()).unwrap().unwrap();
+        } = super::load_auth(codex_home.path(), false).unwrap().unwrap();
         assert_eq!(None, api_key);
         assert_eq!(AuthMode::ChatGPT, mode);
 
         let guard = auth_dot_json.lock().unwrap();
         let auth_dot_json = guard.as_ref().expect("AuthDotJson should exist");
+        let last_refresh = auth_dot_json
+            .last_refresh
+            .expect("last_refresh should be recorded");
+
         assert_eq!(
             &AuthDotJson {
                 openai_api_key: None,
@@ -468,20 +577,17 @@ mod tests {
                     id_token: IdTokenInfo {
                         email: Some("user@example.com".to_string()),
                         chatgpt_plan_type: Some(PlanType::Known(KnownPlan::Pro)),
+                        chatgpt_account_id: None,
                         raw_jwt: fake_jwt,
                     },
                     access_token: "test-access-token".to_string(),
                     refresh_token: "test-refresh-token".to_string(),
                     account_id: None,
                 }),
-                last_refresh: Some(
-                    DateTime::parse_from_rfc3339(LAST_REFRESH)
-                        .unwrap()
-                        .with_timezone(&Utc)
-                ),
+                last_refresh: Some(last_refresh),
             },
             auth_dot_json
-        )
+        );
     }
 
     #[tokio::test]
@@ -494,7 +600,7 @@ mod tests {
         )
         .unwrap();
 
-        let auth = super::load_auth(dir.path()).unwrap().unwrap();
+        let auth = super::load_auth(dir.path(), false).unwrap().unwrap();
         assert_eq!(auth.mode, AuthMode::ApiKey);
         assert_eq!(auth.api_key, Some("sk-test-key".to_string()));
 
@@ -520,6 +626,7 @@ mod tests {
     struct AuthFileParams {
         openai_api_key: Option<String>,
         chatgpt_plan_type: String,
+        chatgpt_account_id: Option<String>,
     }
 
     fn write_auth_file(params: AuthFileParams, codex_home: &Path) -> std::io::Result<String> {
@@ -534,15 +641,21 @@ mod tests {
             alg: "none",
             typ: "JWT",
         };
+        let mut auth_payload = serde_json::json!({
+            "chatgpt_plan_type": params.chatgpt_plan_type,
+            "chatgpt_user_id": "user-12345",
+            "user_id": "user-12345",
+        });
+
+        if let Some(chatgpt_account_id) = params.chatgpt_account_id {
+            let org_value = serde_json::Value::String(chatgpt_account_id);
+            auth_payload["chatgpt_account_id"] = org_value;
+        }
+
         let payload = serde_json::json!({
             "email": "user@example.com",
             "email_verified": true,
-            "https://api.openai.com/auth": {
-                "chatgpt_account_id": "bc3618e3-489d-4d49-9362-1561dc53ba53",
-                "chatgpt_plan_type": params.chatgpt_plan_type,
-                "chatgpt_user_id": "user-12345",
-                "user_id": "user-12345",
-            }
+            "https://api.openai.com/auth": auth_payload,
         });
         let b64 = |b: &[u8]| base64::engine::general_purpose::URL_SAFE_NO_PAD.encode(b);
         let header_b64 = b64(&serde_json::to_vec(&header)?);
@@ -557,12 +670,159 @@ mod tests {
                 "access_token": "test-access-token",
                 "refresh_token": "test-refresh-token"
             },
-            "last_refresh": LAST_REFRESH,
+            "last_refresh": Utc::now(),
         });
         let auth_json = serde_json::to_string_pretty(&auth_json_data)?;
         std::fs::write(auth_file, auth_json)?;
         Ok(fake_jwt)
     }
+
+    fn build_config(
+        codex_home: &Path,
+        forced_login_method: Option<ForcedLoginMethod>,
+        forced_chatgpt_workspace_id: Option<String>,
+    ) -> Config {
+        let mut config = Config::load_from_base_config_with_overrides(
+            ConfigToml::default(),
+            ConfigOverrides::default(),
+            codex_home.to_path_buf(),
+        )
+        .expect("config should load");
+        config.forced_login_method = forced_login_method;
+        config.forced_chatgpt_workspace_id = forced_chatgpt_workspace_id;
+        config
+    }
+
+    /// Use sparingly.
+    /// TODO (gpeal): replace this with an injectable env var provider.
+    #[cfg(test)]
+    struct EnvVarGuard {
+        key: &'static str,
+        original: Option<std::ffi::OsString>,
+    }
+
+    #[cfg(test)]
+    impl EnvVarGuard {
+        fn set(key: &'static str, value: &str) -> Self {
+            let original = env::var_os(key);
+            unsafe {
+                env::set_var(key, value);
+            }
+            Self { key, original }
+        }
+    }
+
+    #[cfg(test)]
+    impl Drop for EnvVarGuard {
+        fn drop(&mut self) {
+            unsafe {
+                match &self.original {
+                    Some(value) => env::set_var(self.key, value),
+                    None => env::remove_var(self.key),
+                }
+            }
+        }
+    }
+
+    #[tokio::test]
+    async fn enforce_login_restrictions_logs_out_for_method_mismatch() {
+        let codex_home = tempdir().unwrap();
+        login_with_api_key(codex_home.path(), "sk-test").expect("seed api key");
+
+        let config = build_config(codex_home.path(), Some(ForcedLoginMethod::Chatgpt), None);
+
+        let err = super::enforce_login_restrictions(&config)
+            .await
+            .expect_err("expected method mismatch to error");
+        assert!(err.to_string().contains("ChatGPT login is required"));
+        assert!(
+            !codex_home.path().join("auth.json").exists(),
+            "auth.json should be removed on mismatch"
+        );
+    }
+
+    #[tokio::test]
+    async fn enforce_login_restrictions_logs_out_for_workspace_mismatch() {
+        let codex_home = tempdir().unwrap();
+        let _jwt = write_auth_file(
+            AuthFileParams {
+                openai_api_key: None,
+                chatgpt_plan_type: "pro".to_string(),
+                chatgpt_account_id: Some("org_another_org".to_string()),
+            },
+            codex_home.path(),
+        )
+        .expect("failed to write auth file");
+
+        let config = build_config(codex_home.path(), None, Some("org_mine".to_string()));
+
+        let err = super::enforce_login_restrictions(&config)
+            .await
+            .expect_err("expected workspace mismatch to error");
+        assert!(err.to_string().contains("workspace org_mine"));
+        assert!(
+            !codex_home.path().join("auth.json").exists(),
+            "auth.json should be removed on mismatch"
+        );
+    }
+
+    #[tokio::test]
+    async fn enforce_login_restrictions_allows_matching_workspace() {
+        let codex_home = tempdir().unwrap();
+        let _jwt = write_auth_file(
+            AuthFileParams {
+                openai_api_key: None,
+                chatgpt_plan_type: "pro".to_string(),
+                chatgpt_account_id: Some("org_mine".to_string()),
+            },
+            codex_home.path(),
+        )
+        .expect("failed to write auth file");
+
+        let config = build_config(codex_home.path(), None, Some("org_mine".to_string()));
+
+        super::enforce_login_restrictions(&config)
+            .await
+            .expect("matching workspace should succeed");
+        assert!(
+            codex_home.path().join("auth.json").exists(),
+            "auth.json should remain when restrictions pass"
+        );
+    }
+
+    #[tokio::test]
+    async fn enforce_login_restrictions_allows_api_key_if_login_method_not_set_but_forced_chatgpt_workspace_id_is_set()
+     {
+        let codex_home = tempdir().unwrap();
+        login_with_api_key(codex_home.path(), "sk-test").expect("seed api key");
+
+        let config = build_config(codex_home.path(), None, Some("org_mine".to_string()));
+
+        super::enforce_login_restrictions(&config)
+            .await
+            .expect("matching workspace should succeed");
+        assert!(
+            codex_home.path().join("auth.json").exists(),
+            "auth.json should remain when restrictions pass"
+        );
+    }
+
+    #[tokio::test]
+    #[serial(codex_api_key)]
+    async fn enforce_login_restrictions_blocks_env_api_key_when_chatgpt_required() {
+        let _guard = EnvVarGuard::set(CODEX_API_KEY_ENV_VAR, "sk-env");
+        let codex_home = tempdir().unwrap();
+
+        let config = build_config(codex_home.path(), Some(ForcedLoginMethod::Chatgpt), None);
+
+        let err = super::enforce_login_restrictions(&config)
+            .await
+            .expect_err("environment API key should not satisfy forced ChatGPT login");
+        assert!(
+            err.to_string()
+                .contains("ChatGPT login is required, but an API key is currently being used.")
+        );
+    }
 }
 
 /// Central manager providing a single source of truth for auth.json derived
@@ -577,6 +837,7 @@ mod tests {
 pub struct AuthManager {
     codex_home: PathBuf,
     inner: RwLock<CachedAuth>,
+    enable_codex_api_key_env: bool,
 }
 
 impl AuthManager {
@@ -584,11 +845,14 @@ impl AuthManager {
     /// preferred auth method. Errors loading auth are swallowed; `auth()` will
     /// simply return `None` in that case so callers can treat it as an
     /// unauthenticated state.
-    pub fn new(codex_home: PathBuf) -> Self {
-        let auth = CodexAuth::from_codex_home(&codex_home).ok().flatten();
+    pub fn new(codex_home: PathBuf, enable_codex_api_key_env: bool) -> Self {
+        let auth = load_auth(&codex_home, enable_codex_api_key_env)
+            .ok()
+            .flatten();
         Self {
             codex_home,
             inner: RwLock::new(CachedAuth { auth }),
+            enable_codex_api_key_env,
         }
     }
 
@@ -598,6 +862,7 @@ impl AuthManager {
         Arc::new(Self {
             codex_home: PathBuf::new(),
             inner: RwLock::new(cached),
+            enable_codex_api_key_env: false,
         })
     }
 
@@ -609,7 +874,9 @@ impl AuthManager {
     /// Force a reload of the auth information from auth.json. Returns
     /// whether the auth value changed.
     pub fn reload(&self) -> bool {
-        let new_auth = CodexAuth::from_codex_home(&self.codex_home).ok().flatten();
+        let new_auth = load_auth(&self.codex_home, self.enable_codex_api_key_env)
+            .ok()
+            .flatten();
         if let Ok(mut guard) = self.inner.write() {
             let changed = !AuthManager::auths_equal(&guard.auth, &new_auth);
             guard.auth = new_auth;
@@ -628,8 +895,8 @@ impl AuthManager {
     }
 
     /// Convenience constructor returning an `Arc` wrapper.
-    pub fn shared(codex_home: PathBuf) -> Arc<Self> {
-        Arc::new(Self::new(codex_home))
+    pub fn shared(codex_home: PathBuf, enable_codex_api_key_env: bool) -> Arc<Self> {
+        Arc::new(Self::new(codex_home, enable_codex_api_key_env))
     }
 
     /// Attempt to refresh the current auth token (if any). On success, reload

codex-rs/core/src/chat_completions.rs

@@ -5,11 +5,13 @@ use crate::client_common::Prompt;
 use crate::client_common::ResponseEvent;
 use crate::client_common::ResponseStream;
 use crate::error::CodexErr;
+use crate::error::ConnectionFailedError;
+use crate::error::ResponseStreamFailed;
 use crate::error::Result;
 use crate::error::RetryLimitReachedError;
 use crate::error::UnexpectedResponseError;
 use crate::model_family::ModelFamily;
-use crate::openai_tools::create_tools_json_for_chat_completions_api;
+use crate::tools::spec::create_tools_json_for_chat_completions_api;
 use crate::util::backoff;
 use bytes::Bytes;
 use codex_otel::otel_event_manager::OtelEventManager;
@@ -309,7 +311,12 @@ pub(crate) async fn stream_chat_completions(
         match res {
             Ok(resp) if resp.status().is_success() => {
                 let (tx_event, rx_event) = mpsc::channel::<Result<ResponseEvent>>(1600);
-                let stream = resp.bytes_stream().map_err(CodexErr::Reqwest);
+                let stream = resp.bytes_stream().map_err(|e| {
+                    CodexErr::ResponseStreamFailed(ResponseStreamFailed {
+                        source: e,
+                        request_id: None,
+                    })
+                });
                 tokio::spawn(process_chat_sse(
                     stream,
                     tx_event,
@@ -349,7 +356,9 @@ pub(crate) async fn stream_chat_completions(
             }
             Err(e) => {
                 if attempt > max_retries {
-                    return Err(e.into());
+                    return Err(CodexErr::ConnectionFailed(ConnectionFailedError {
+                        source: e,
+                    }));
                 }
                 let delay = backoff(attempt);
                 tokio::time::sleep(delay).await;
@@ -389,10 +398,12 @@ async fn process_chat_sse<S>(
     let mut reasoning_text = String::new();
 
     loop {
-        let sse = match otel_event_manager
-            .log_sse_event(|| timeout(idle_timeout, stream.next()))
-            .await
-        {
+        let start = std::time::Instant::now();
+        let response = timeout(idle_timeout, stream.next()).await;
+        let duration = start.elapsed();
+        otel_event_manager.log_sse_event(&response, duration);
+
+        let sse = match response {
             Ok(Some(Ok(ev))) => ev,
             Ok(Some(Err(e))) => {
                 let _ = tx_event

codex-rs/core/src/client.rs

@@ -5,6 +5,8 @@ use std::time::Duration;
 
 use crate::AuthManager;
 use crate::auth::CodexAuth;
+use crate::error::ConnectionFailedError;
+use crate::error::ResponseStreamFailed;
 use crate::error::RetryLimitReachedError;
 use crate::error::UnexpectedResponseError;
 use bytes::Bytes;
@@ -43,12 +45,15 @@ use crate::model_family::ModelFamily;
 use crate::model_provider_info::ModelProviderInfo;
 use crate::model_provider_info::WireApi;
 use crate::openai_model_info::get_model_info;
-use crate::openai_tools::create_tools_json_for_responses_api;
 use crate::protocol::RateLimitSnapshot;
 use crate::protocol::RateLimitWindow;
 use crate::protocol::TokenUsage;
+use crate::state::TaskKind;
 use crate::token_data::PlanType;
+use crate::tools::spec::create_tools_json_for_responses_api;
 use crate::util::backoff;
+use chrono::DateTime;
+use chrono::Utc;
 use codex_otel::otel_event_manager::OtelEventManager;
 use codex_protocol::config_types::ReasoningEffort as ReasoningEffortConfig;
 use codex_protocol::config_types::ReasoningSummary as ReasoningSummaryConfig;
@@ -63,13 +68,12 @@ struct ErrorResponse {
 #[derive(Debug, Deserialize)]
 struct Error {
     r#type: Option<String>,
-    #[allow(dead_code)]
     code: Option<String>,
     message: Option<String>,
 
     // Optional fields available on "usage_limit_reached" and "usage_not_included" errors
     plan_type: Option<PlanType>,
-    resets_in_seconds: Option<u64>,
+    resets_at: Option<i64>,
 }
 
 #[derive(Debug, Clone)]
@@ -108,10 +112,12 @@ impl ModelClient {
         }
     }
 
-    pub fn get_model_context_window(&self) -> Option<u64> {
+    pub fn get_model_context_window(&self) -> Option<i64> {
+        let pct = self.config.model_family.effective_context_window_percent;
         self.config
             .model_context_window
             .or_else(|| get_model_info(&self.config.model_family).map(|info| info.context_window))
+            .map(|w| w.saturating_mul(pct) / 100)
     }
 
     pub fn get_auto_compact_token_limit(&self) -> Option<i64> {
@@ -124,8 +130,16 @@ impl ModelClient {
     /// the provider config.  Public callers always invoke `stream()` – the
     /// specialised helpers are private to avoid accidental misuse.
     pub async fn stream(&self, prompt: &Prompt) -> Result<ResponseStream> {
+        self.stream_with_task_kind(prompt, TaskKind::Regular).await
+    }
+
+    pub(crate) async fn stream_with_task_kind(
+        &self,
+        prompt: &Prompt,
+        task_kind: TaskKind,
+    ) -> Result<ResponseStream> {
         match self.provider.wire_api {
-            WireApi::Responses => self.stream_responses(prompt).await,
+            WireApi::Responses => self.stream_responses(prompt, task_kind).await,
             WireApi::Chat => {
                 // Create the raw streaming connection first.
                 let response_stream = stream_chat_completions(
@@ -166,7 +180,11 @@ impl ModelClient {
     }
 
     /// Implementation for the OpenAI *Responses* experimental API.
-    async fn stream_responses(&self, prompt: &Prompt) -> Result<ResponseStream> {
+    async fn stream_responses(
+        &self,
+        prompt: &Prompt,
+        task_kind: TaskKind,
+    ) -> Result<ResponseStream> {
         if let Some(path) = &*CODEX_RS_SSE_FIXTURE {
             // short circuit for tests
             warn!(path, "Streaming from fixture");
@@ -228,7 +246,7 @@ impl ModelClient {
             input: &input_with_instructions,
             tools: &tools_json,
             tool_choice: "auto",
-            parallel_tool_calls: false,
+            parallel_tool_calls: prompt.parallel_tool_calls,
             reasoning,
             store: azure_workaround,
             stream: true,
@@ -245,7 +263,7 @@ impl ModelClient {
         let max_attempts = self.provider.request_max_retries();
         for attempt in 0..=max_attempts {
             match self
-                .attempt_stream_responses(attempt, &payload_json, &auth_manager)
+                .attempt_stream_responses(attempt, &payload_json, &auth_manager, task_kind)
                 .await
             {
                 Ok(stream) => {
@@ -273,6 +291,7 @@ impl ModelClient {
         attempt: u64,
         payload_json: &Value,
         auth_manager: &Option<Arc<AuthManager>>,
+        task_kind: TaskKind,
     ) -> std::result::Result<ResponseStream, StreamAttemptError> {
         // Always fetch the latest auth in case a prior attempt refreshed the token.
         let auth = auth_manager.as_ref().and_then(|m| m.auth());
@@ -295,6 +314,7 @@ impl ModelClient {
             .header("conversation_id", self.conversation_id.to_string())
             .header("session_id", self.conversation_id.to_string())
             .header(reqwest::header::ACCEPT, "text/event-stream")
+            .header("Codex-Task-Type", task_kind.header_value())
             .json(payload_json);
 
         if let Some(auth) = auth.as_ref()
@@ -337,7 +357,12 @@ impl ModelClient {
                 }
 
                 // spawn task to process SSE
-                let stream = resp.bytes_stream().map_err(CodexErr::Reqwest);
+                let stream = resp.bytes_stream().map_err(move |e| {
+                    CodexErr::ResponseStreamFailed(ResponseStreamFailed {
+                        source: e,
+                        request_id: request_id.clone(),
+                    })
+                });
                 tokio::spawn(process_sse(
                     stream,
                     tx_event,
@@ -398,10 +423,12 @@ impl ModelClient {
                             let plan_type = error
                                 .plan_type
                                 .or_else(|| auth.as_ref().and_then(CodexAuth::get_plan_type));
-                            let resets_in_seconds = error.resets_in_seconds;
+                            let resets_at = error
+                                .resets_at
+                                .and_then(|seconds| DateTime::<Utc>::from_timestamp(seconds, 0));
                             let codex_err = CodexErr::UsageLimitReached(UsageLimitReachedError {
                                 plan_type,
-                                resets_in_seconds,
+                                resets_at,
                                 rate_limits: rate_limit_snapshot,
                             });
                             return Err(StreamAttemptError::Fatal(codex_err));
@@ -417,7 +444,9 @@ impl ModelClient {
                     request_id,
                 })
             }
-            Err(e) => Err(StreamAttemptError::RetryableTransportError(e.into())),
+            Err(e) => Err(StreamAttemptError::RetryableTransportError(
+                CodexErr::ConnectionFailed(ConnectionFailedError { source: e }),
+            )),
         }
     }
 
@@ -515,11 +544,11 @@ struct ResponseCompleted {
 
 #[derive(Debug, Deserialize)]
 struct ResponseCompletedUsage {
-    input_tokens: u64,
+    input_tokens: i64,
     input_tokens_details: Option<ResponseCompletedInputTokensDetails>,
-    output_tokens: u64,
+    output_tokens: i64,
     output_tokens_details: Option<ResponseCompletedOutputTokensDetails>,
-    total_tokens: u64,
+    total_tokens: i64,
 }
 
 impl From<ResponseCompletedUsage> for TokenUsage {
@@ -542,12 +571,12 @@ impl From<ResponseCompletedUsage> for TokenUsage {
 
 #[derive(Debug, Deserialize)]
 struct ResponseCompletedInputTokensDetails {
-    cached_tokens: u64,
+    cached_tokens: i64,
 }
 
 #[derive(Debug, Deserialize)]
 struct ResponseCompletedOutputTokensDetails {
-    reasoning_tokens: u64,
+    reasoning_tokens: i64,
 }
 
 fn attach_item_ids(payload_json: &mut Value, original_items: &[ResponseItem]) {
@@ -582,14 +611,14 @@ fn parse_rate_limit_snapshot(headers: &HeaderMap) -> Option<RateLimitSnapshot> {
         headers,
         "x-codex-primary-used-percent",
         "x-codex-primary-window-minutes",
-        "x-codex-primary-reset-after-seconds",
+        "x-codex-primary-reset-at",
     );
 
     let secondary = parse_rate_limit_window(
         headers,
         "x-codex-secondary-used-percent",
         "x-codex-secondary-window-minutes",
-        "x-codex-secondary-reset-after-seconds",
+        "x-codex-secondary-reset-at",
     );
 
     Some(RateLimitSnapshot { primary, secondary })
@@ -604,17 +633,17 @@ fn parse_rate_limit_window(
     let used_percent: Option<f64> = parse_header_f64(headers, used_percent_header);
 
     used_percent.and_then(|used_percent| {
-        let window_minutes = parse_header_u64(headers, window_minutes_header);
-        let resets_in_seconds = parse_header_u64(headers, resets_header);
+        let window_minutes = parse_header_i64(headers, window_minutes_header);
+        let resets_at = parse_header_i64(headers, resets_header);
 
         let has_data = used_percent != 0.0
             || window_minutes.is_some_and(|minutes| minutes != 0)
-            || resets_in_seconds.is_some_and(|seconds| seconds != 0);
+            || resets_at.is_some();
 
         has_data.then_some(RateLimitWindow {
             used_percent,
             window_minutes,
-            resets_in_seconds,
+            resets_at,
         })
     })
 }
@@ -626,8 +655,8 @@ fn parse_header_f64(headers: &HeaderMap, name: &str) -> Option<f64> {
         .filter(|v| v.is_finite())
 }
 
-fn parse_header_u64(headers: &HeaderMap, name: &str) -> Option<u64> {
-    parse_header_str(headers, name)?.parse::<u64>().ok()
+fn parse_header_i64(headers: &HeaderMap, name: &str) -> Option<i64> {
+    parse_header_str(headers, name)?.parse::<i64>().ok()
 }
 
 fn parse_header_str<'a>(headers: &'a HeaderMap, name: &str) -> Option<&'a str> {
@@ -650,10 +679,12 @@ async fn process_sse<S>(
     let mut response_error: Option<CodexErr> = None;
 
     loop {
-        let sse = match otel_event_manager
-            .log_sse_event(|| timeout(idle_timeout, stream.next()))
-            .await
-        {
+        let start = std::time::Instant::now();
+        let response = timeout(idle_timeout, stream.next()).await;
+        let duration = start.elapsed();
+        otel_event_manager.log_sse_event(&response, duration);
+
+        let sse = match response {
             Ok(Some(Ok(sse))) => sse,
             Ok(Some(Err(e))) => {
                 debug!("SSE Error: {e:#}");
@@ -794,9 +825,13 @@ async fn process_sse<S>(
                     if let Some(error) = error {
                         match serde_json::from_value::<Error>(error.clone()) {
                             Ok(error) => {
-                                let delay = try_parse_retry_after(&error);
-                                let message = error.message.unwrap_or_default();
-                                response_error = Some(CodexErr::Stream(message, delay));
+                                if is_context_window_error(&error) {
+                                    response_error = Some(CodexErr::ContextWindowExceeded);
+                                } else {
+                                    let delay = try_parse_retry_after(&error);
+                                    let message = error.message.clone().unwrap_or_default();
+                                    response_error = Some(CodexErr::Stream(message, delay));
+                                }
                             }
                             Err(e) => {
                                 let error = format!("failed to parse ErrorResponse: {e}");
@@ -922,9 +957,14 @@ fn try_parse_retry_after(err: &Error) -> Option<Duration> {
     None
 }
 
+fn is_context_window_error(error: &Error) -> bool {
+    error.code.as_deref() == Some("context_length_exceeded")
+}
+
 #[cfg(test)]
 mod tests {
     use super::*;
+    use assert_matches::assert_matches;
     use serde_json::json;
     use tokio::sync::mpsc;
     use tokio_test::io::Builder as IoBuilder;
@@ -1005,6 +1045,7 @@ mod tests {
             "test",
             "test",
             None,
+            Some("test@test.com".to_string()),
             Some(AuthMode::ChatGPT),
             false,
             "test".to_string(),
@@ -1179,6 +1220,74 @@ mod tests {
         }
     }
 
+    #[tokio::test]
+    async fn context_window_error_is_fatal() {
+        let raw_error = r#"{"type":"response.failed","sequence_number":3,"response":{"id":"resp_5c66275b97b9baef1ed95550adb3b7ec13b17aafd1d2f11b","object":"response","created_at":1759510079,"status":"failed","background":false,"error":{"code":"context_length_exceeded","message":"Your input exceeds the context window of this model. Please adjust your input and try again."},"usage":null,"user":null,"metadata":{}}}"#;
+
+        let sse1 = format!("event: response.failed\ndata: {raw_error}\n\n");
+        let provider = ModelProviderInfo {
+            name: "test".to_string(),
+            base_url: Some("https://test.com".to_string()),
+            env_key: Some("TEST_API_KEY".to_string()),
+            env_key_instructions: None,
+            wire_api: WireApi::Responses,
+            query_params: None,
+            http_headers: None,
+            env_http_headers: None,
+            request_max_retries: Some(0),
+            stream_max_retries: Some(0),
+            stream_idle_timeout_ms: Some(1000),
+            requires_openai_auth: false,
+        };
+
+        let otel_event_manager = otel_event_manager();
+
+        let events = collect_events(&[sse1.as_bytes()], provider, otel_event_manager).await;
+
+        assert_eq!(events.len(), 1);
+
+        match &events[0] {
+            Err(err @ CodexErr::ContextWindowExceeded) => {
+                assert_eq!(err.to_string(), CodexErr::ContextWindowExceeded.to_string());
+            }
+            other => panic!("unexpected context window event: {other:?}"),
+        }
+    }
+
+    #[tokio::test]
+    async fn context_window_error_with_newline_is_fatal() {
+        let raw_error = r#"{"type":"response.failed","sequence_number":4,"response":{"id":"resp_fatal_newline","object":"response","created_at":1759510080,"status":"failed","background":false,"error":{"code":"context_length_exceeded","message":"Your input exceeds the context window of this model. Please adjust your input and try\nagain."},"usage":null,"user":null,"metadata":{}}}"#;
+
+        let sse1 = format!("event: response.failed\ndata: {raw_error}\n\n");
+        let provider = ModelProviderInfo {
+            name: "test".to_string(),
+            base_url: Some("https://test.com".to_string()),
+            env_key: Some("TEST_API_KEY".to_string()),
+            env_key_instructions: None,
+            wire_api: WireApi::Responses,
+            query_params: None,
+            http_headers: None,
+            env_http_headers: None,
+            request_max_retries: Some(0),
+            stream_max_retries: Some(0),
+            stream_idle_timeout_ms: Some(1000),
+            requires_openai_auth: false,
+        };
+
+        let otel_event_manager = otel_event_manager();
+
+        let events = collect_events(&[sse1.as_bytes()], provider, otel_event_manager).await;
+
+        assert_eq!(events.len(), 1);
+
+        match &events[0] {
+            Err(err @ CodexErr::ContextWindowExceeded) => {
+                assert_eq!(err.to_string(), CodexErr::ContextWindowExceeded.to_string());
+            }
+            other => panic!("unexpected context window event: {other:?}"),
+        }
+    }
+
     // ────────────────────────────
     // Table-driven test from `main`
     // ────────────────────────────
@@ -1287,7 +1396,7 @@ mod tests {
             message: Some("Rate limit reached for gpt-5 in organization org- on tokens per min (TPM): Limit 1, Used 1, Requested 19304. Please try again in 28ms. Visit https://platform.openai.com/account/rate-limits to learn more.".to_string()),
             code: Some("rate_limit_exceeded".to_string()),
             plan_type: None,
-            resets_in_seconds: None
+            resets_at: None
         };
 
         let delay = try_parse_retry_after(&err);
@@ -1301,40 +1410,36 @@ mod tests {
             message: Some("Rate limit reached for gpt-5 in organization <ORG> on tokens per min (TPM): Limit 30000, Used 6899, Requested 24050. Please try again in 1.898s. Visit https://platform.openai.com/account/rate-limits to learn more.".to_string()),
             code: Some("rate_limit_exceeded".to_string()),
             plan_type: None,
-            resets_in_seconds: None
+            resets_at: None
         };
         let delay = try_parse_retry_after(&err);
         assert_eq!(delay, Some(Duration::from_secs_f64(1.898)));
     }
 
     #[test]
-    fn error_response_deserializes_old_schema_known_plan_type_and_serializes_back() {
+    fn error_response_deserializes_schema_known_plan_type_and_serializes_back() {
         use crate::token_data::KnownPlan;
         use crate::token_data::PlanType;
 
-        let json = r#"{"error":{"type":"usage_limit_reached","plan_type":"pro","resets_in_seconds":3600}}"#;
-        let resp: ErrorResponse =
-            serde_json::from_str(json).expect("should deserialize old schema");
+        let json =
+            r#"{"error":{"type":"usage_limit_reached","plan_type":"pro","resets_at":1704067200}}"#;
+        let resp: ErrorResponse = serde_json::from_str(json).expect("should deserialize schema");
 
-        assert!(matches!(
-            resp.error.plan_type,
-            Some(PlanType::Known(KnownPlan::Pro))
-        ));
+        assert_matches!(resp.error.plan_type, Some(PlanType::Known(KnownPlan::Pro)));
 
         let plan_json = serde_json::to_string(&resp.error.plan_type).expect("serialize plan_type");
         assert_eq!(plan_json, "\"pro\"");
     }
 
     #[test]
-    fn error_response_deserializes_old_schema_unknown_plan_type_and_serializes_back() {
+    fn error_response_deserializes_schema_unknown_plan_type_and_serializes_back() {
         use crate::token_data::PlanType;
 
         let json =
-            r#"{"error":{"type":"usage_limit_reached","plan_type":"vip","resets_in_seconds":60}}"#;
-        let resp: ErrorResponse =
-            serde_json::from_str(json).expect("should deserialize old schema");
+            r#"{"error":{"type":"usage_limit_reached","plan_type":"vip","resets_at":1704067260}}"#;
+        let resp: ErrorResponse = serde_json::from_str(json).expect("should deserialize schema");
 
-        assert!(matches!(resp.error.plan_type, Some(PlanType::Unknown(ref s)) if s == "vip"));
+        assert_matches!(resp.error.plan_type, Some(PlanType::Unknown(ref s)) if s == "vip");
 
         let plan_json = serde_json::to_string(&resp.error.plan_type).expect("serialize plan_type");
         assert_eq!(plan_json, "\"vip\"");

codex-rs/core/src/client_common.rs

@@ -1,6 +1,6 @@
+use crate::client_common::tools::ToolSpec;
 use crate::error::Result;
 use crate::model_family::ModelFamily;
-use crate::openai_tools::OpenAiTool;
 use crate::protocol::RateLimitSnapshot;
 use crate::protocol::TokenUsage;
 use codex_apply_patch::APPLY_PATCH_TOOL_INSTRUCTIONS;
@@ -9,9 +9,11 @@ use codex_protocol::config_types::ReasoningSummary as ReasoningSummaryConfig;
 use codex_protocol::config_types::Verbosity as VerbosityConfig;
 use codex_protocol::models::ResponseItem;
 use futures::Stream;
+use serde::Deserialize;
 use serde::Serialize;
 use serde_json::Value;
 use std::borrow::Cow;
+use std::collections::HashSet;
 use std::ops::Deref;
 use std::pin::Pin;
 use std::task::Context;
@@ -29,7 +31,10 @@ pub struct Prompt {
 
     /// Tools available to the model, including additional tools sourced from
     /// external MCP servers.
-    pub(crate) tools: Vec<OpenAiTool>,
+    pub(crate) tools: Vec<ToolSpec>,
+
+    /// Whether parallel tool calls are permitted for this prompt.
+    pub(crate) parallel_tool_calls: bool,
 
     /// Optional override for the built-in BASE_INSTRUCTIONS.
     pub base_instructions_override: Option<String>,
@@ -49,8 +54,8 @@ impl Prompt {
         // AND
         // - there is no apply_patch tool present
         let is_apply_patch_tool_present = self.tools.iter().any(|tool| match tool {
-            OpenAiTool::Function(f) => f.name == "apply_patch",
-            OpenAiTool::Freeform(f) => f.name == "apply_patch",
+            ToolSpec::Function(f) => f.name == "apply_patch",
+            ToolSpec::Freeform(f) => f.name == "apply_patch",
             _ => false,
         });
         if self.base_instructions_override.is_none()
@@ -64,10 +69,125 @@ impl Prompt {
     }
 
     pub(crate) fn get_formatted_input(&self) -> Vec<ResponseItem> {
-        self.input.clone()
+        let mut input = self.input.clone();
+
+        // when using the *Freeform* apply_patch tool specifically, tool outputs
+        // should be structured text, not json. Do NOT reserialize when using
+        // the Function tool - note that this differs from the check above for
+        // instructions. We declare the result as a named variable for clarity.
+        let is_freeform_apply_patch_tool_present = self.tools.iter().any(|tool| match tool {
+            ToolSpec::Freeform(f) => f.name == "apply_patch",
+            _ => false,
+        });
+        if is_freeform_apply_patch_tool_present {
+            reserialize_shell_outputs(&mut input);
+        }
+
+        input
     }
 }
 
+fn reserialize_shell_outputs(items: &mut [ResponseItem]) {
+    let mut shell_call_ids: HashSet<String> = HashSet::new();
+
+    items.iter_mut().for_each(|item| match item {
+        ResponseItem::LocalShellCall { call_id, id, .. } => {
+            if let Some(identifier) = call_id.clone().or_else(|| id.clone()) {
+                shell_call_ids.insert(identifier);
+            }
+        }
+        ResponseItem::CustomToolCall {
+            id: _,
+            status: _,
+            call_id,
+            name,
+            input: _,
+        } => {
+            if name == "apply_patch" {
+                shell_call_ids.insert(call_id.clone());
+            }
+        }
+        ResponseItem::CustomToolCallOutput { call_id, output } => {
+            if shell_call_ids.remove(call_id)
+                && let Some(structured) = parse_structured_shell_output(output)
+            {
+                *output = structured
+            }
+        }
+        ResponseItem::FunctionCall { name, call_id, .. }
+            if is_shell_tool_name(name) || name == "apply_patch" =>
+        {
+            shell_call_ids.insert(call_id.clone());
+        }
+        ResponseItem::FunctionCallOutput { call_id, output } => {
+            if shell_call_ids.remove(call_id)
+                && let Some(structured) = parse_structured_shell_output(&output.content)
+            {
+                output.content = structured
+            }
+        }
+        _ => {}
+    })
+}
+
+fn is_shell_tool_name(name: &str) -> bool {
+    matches!(name, "shell" | "container.exec")
+}
+
+#[derive(Deserialize)]
+struct ExecOutputJson {
+    output: String,
+    metadata: ExecOutputMetadataJson,
+}
+
+#[derive(Deserialize)]
+struct ExecOutputMetadataJson {
+    exit_code: i32,
+    duration_seconds: f32,
+}
+
+fn parse_structured_shell_output(raw: &str) -> Option<String> {
+    let parsed: ExecOutputJson = serde_json::from_str(raw).ok()?;
+    Some(build_structured_output(&parsed))
+}
+
+fn build_structured_output(parsed: &ExecOutputJson) -> String {
+    let mut sections = Vec::new();
+    sections.push(format!("Exit code: {}", parsed.metadata.exit_code));
+    sections.push(format!(
+        "Wall time: {} seconds",
+        parsed.metadata.duration_seconds
+    ));
+
+    let mut output = parsed.output.clone();
+    if let Some(total_lines) = extract_total_output_lines(&parsed.output) {
+        sections.push(format!("Total output lines: {total_lines}"));
+        if let Some(stripped) = strip_total_output_header(&output) {
+            output = stripped.to_string();
+        }
+    }
+
+    sections.push("Output:".to_string());
+    sections.push(output);
+
+    sections.join("\n")
+}
+
+fn extract_total_output_lines(output: &str) -> Option<u32> {
+    let marker_start = output.find("[... omitted ")?;
+    let marker = &output[marker_start..];
+    let (_, after_of) = marker.split_once(" of ")?;
+    let (total_segment, _) = after_of.split_once(' ')?;
+    total_segment.parse::<u32>().ok()
+}
+
+fn strip_total_output_header(output: &str) -> Option<&str> {
+    let after_prefix = output.strip_prefix("Total output lines: ")?;
+    let (_, remainder) = after_prefix.split_once('\n')?;
+    let remainder = remainder.strip_prefix('\n').unwrap_or(remainder);
+    Some(remainder)
+}
+
 #[derive(Debug)]
 pub enum ResponseEvent {
     Created,
@@ -160,6 +280,65 @@ pub(crate) struct ResponsesApiRequest<'a> {
     pub(crate) text: Option<TextControls>,
 }
 
+pub(crate) mod tools {
+    use crate::tools::spec::JsonSchema;
+    use serde::Deserialize;
+    use serde::Serialize;
+
+    /// When serialized as JSON, this produces a valid "Tool" in the OpenAI
+    /// Responses API.
+    #[derive(Debug, Clone, Serialize, PartialEq)]
+    #[serde(tag = "type")]
+    pub(crate) enum ToolSpec {
+        #[serde(rename = "function")]
+        Function(ResponsesApiTool),
+        #[serde(rename = "local_shell")]
+        LocalShell {},
+        // TODO: Understand why we get an error on web_search although the API docs say it's supported.
+        // https://platform.openai.com/docs/guides/tools-web-search?api-mode=responses#:~:text=%7B%20type%3A%20%22web_search%22%20%7D%2C
+        #[serde(rename = "web_search")]
+        WebSearch {},
+        #[serde(rename = "custom")]
+        Freeform(FreeformTool),
+    }
+
+    impl ToolSpec {
+        pub(crate) fn name(&self) -> &str {
+            match self {
+                ToolSpec::Function(tool) => tool.name.as_str(),
+                ToolSpec::LocalShell {} => "local_shell",
+                ToolSpec::WebSearch {} => "web_search",
+                ToolSpec::Freeform(tool) => tool.name.as_str(),
+            }
+        }
+    }
+
+    #[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
+    pub struct FreeformTool {
+        pub(crate) name: String,
+        pub(crate) description: String,
+        pub(crate) format: FreeformToolFormat,
+    }
+
+    #[derive(Debug, Clone, Serialize, Deserialize, PartialEq)]
+    pub struct FreeformToolFormat {
+        pub(crate) r#type: String,
+        pub(crate) syntax: String,
+        pub(crate) definition: String,
+    }
+
+    #[derive(Debug, Clone, Serialize, PartialEq)]
+    pub struct ResponsesApiTool {
+        pub(crate) name: String,
+        pub(crate) description: String,
+        /// TODO: Validation. When strict is set to true, the JSON schema,
+        /// `required` and `additional_properties` must be present. All fields in
+        /// `properties` must be present in `required`.
+        pub(crate) strict: bool,
+        pub(crate) parameters: JsonSchema,
+    }
+}
+
 pub(crate) fn create_reasoning_param_for_request(
     model_family: &ModelFamily,
     effort: Option<ReasoningEffortConfig>,
@@ -279,7 +458,7 @@ mod tests {
             input: &input,
             tools: &tools,
             tool_choice: "auto",
-            parallel_tool_calls: false,
+            parallel_tool_calls: true,
             reasoning: None,
             store: false,
             stream: true,
@@ -320,7 +499,7 @@ mod tests {
             input: &input,
             tools: &tools,
             tool_choice: "auto",
-            parallel_tool_calls: false,
+            parallel_tool_calls: true,
             reasoning: None,
             store: false,
             stream: true,
@@ -356,7 +535,7 @@ mod tests {
             input: &input,
             tools: &tools,
             tool_choice: "auto",
-            parallel_tool_calls: false,
+            parallel_tool_calls: true,
             reasoning: None,
             store: false,
             stream: true,

codex-rs/core/src/codex/compact.rs

@@ -16,6 +16,7 @@ use crate::protocol::InputItem;
 use crate::protocol::InputMessageKind;
 use crate::protocol::TaskStartedEvent;
 use crate::protocol::TurnContextItem;
+use crate::state::TaskKind;
 use crate::truncate::truncate_middle;
 use crate::util::backoff;
 use askama::Template;
@@ -70,14 +71,10 @@ async fn run_compact_task_inner(
     input: Vec<InputItem>,
 ) {
     let initial_input_for_turn: ResponseInputItem = ResponseInputItem::from(input);
-    let turn_input = sess
+    let mut turn_input = sess
         .turn_input_with_history(vec![initial_input_for_turn.clone().into()])
         .await;
-
-    let prompt = Prompt {
-        input: turn_input,
-        ..Default::default()
-    };
+    let mut truncated_count = 0usize;
 
     let max_retries = turn_context.client.get_provider().stream_max_retries();
     let mut retries = 0;
@@ -93,25 +90,54 @@ async fn run_compact_task_inner(
     sess.persist_rollout_items(&[rollout_item]).await;
 
     loop {
+        let prompt = Prompt {
+            input: turn_input.clone(),
+            ..Default::default()
+        };
         let attempt_result =
             drain_to_completed(&sess, turn_context.as_ref(), &sub_id, &prompt).await;
 
         match attempt_result {
             Ok(()) => {
+                if truncated_count > 0 {
+                    sess.notify_background_event(
+                        &sub_id,
+                        format!(
+                            "Trimmed {truncated_count} older conversation item(s) before compacting so the prompt fits the model context window."
+                        ),
+                    )
+                    .await;
+                }
                 break;
             }
             Err(CodexErr::Interrupted) => {
                 return;
             }
+            Err(e @ CodexErr::ContextWindowExceeded) => {
+                if turn_input.len() > 1 {
+                    turn_input.remove(0);
+                    truncated_count += 1;
+                    retries = 0;
+                    continue;
+                }
+                sess.set_total_tokens_full(&sub_id, turn_context.as_ref())
+                    .await;
+                let event = Event {
+                    id: sub_id.clone(),
+                    msg: EventMsg::Error(ErrorEvent {
+                        message: e.to_string(),
+                    }),
+                };
+                sess.send_event(event).await;
+                return;
+            }
             Err(e) => {
                 if retries < max_retries {
                     retries += 1;
                     let delay = backoff(retries);
                     sess.notify_stream_error(
                         &sub_id,
-                        format!(
-                            "stream error: {e}; retrying {retries}/{max_retries} in {delay:?}…"
-                        ),
+                        format!("Re-connecting... {retries}/{max_retries}"),
                     )
                     .await;
                     tokio::time::sleep(delay).await;
@@ -233,7 +259,11 @@ async fn drain_to_completed(
     sub_id: &str,
     prompt: &Prompt,
 ) -> CodexResult<()> {
-    let mut stream = turn_context.client.clone().stream(prompt).await?;
+    let mut stream = turn_context
+        .client
+        .clone()
+        .stream_with_task_kind(prompt, TaskKind::Compact)
+        .await?;
     loop {
         let maybe_event = stream.next().await;
         let Some(event) = maybe_event else {

codex-rs/core/src/command_safety/is_safe_command.rs

@@ -1,15 +1,25 @@
 use crate::bash::parse_bash_lc_plain_commands;
 
 pub fn is_known_safe_command(command: &[String]) -> bool {
+    let command: Vec<String> = command
+        .iter()
+        .map(|s| {
+            if s == "zsh" {
+                "bash".to_string()
+            } else {
+                s.clone()
+            }
+        })
+        .collect();
     #[cfg(target_os = "windows")]
     {
         use super::windows_safe_commands::is_safe_command_windows;
-        if is_safe_command_windows(command) {
+        if is_safe_command_windows(&command) {
             return true;
         }
     }
 
-    if is_safe_to_call_with_exec(command) {
+    if is_safe_to_call_with_exec(&command) {
         return true;
     }
 
@@ -19,7 +29,7 @@ pub fn is_known_safe_command(command: &[String]) -> bool {
     // introduce side effects ( "&&", "||", ";", and "|" ). If every
     // individual command in the script is itself a known‑safe command, then
     // the composite expression is considered safe.
-    if let Some(all_commands) = parse_bash_lc_plain_commands(command)
+    if let Some(all_commands) = parse_bash_lc_plain_commands(&command)
         && !all_commands.is_empty()
         && all_commands
             .iter()
@@ -31,9 +41,14 @@ pub fn is_known_safe_command(command: &[String]) -> bool {
 }
 
 fn is_safe_to_call_with_exec(command: &[String]) -> bool {
-    let cmd0 = command.first().map(String::as_str);
+    let Some(cmd0) = command.first().map(String::as_str) else {
+        return false;
+    };
 
-    match cmd0 {
+    match std::path::Path::new(&cmd0)
+        .file_name()
+        .and_then(|osstr| osstr.to_str())
+    {
         #[rustfmt::skip]
         Some(
             "cat" |
@@ -103,13 +118,12 @@ fn is_safe_to_call_with_exec(command: &[String]) -> bool {
         // Rust
         Some("cargo") if command.get(1).map(String::as_str) == Some("check") => true,
 
-        // Special-case `sed -n {N|M,N}p FILE`
+        // Special-case `sed -n {N|M,N}p`
         Some("sed")
             if {
-                command.len() == 4
+                command.len() <= 4
                     && command.get(1).map(String::as_str) == Some("-n")
                     && is_valid_sed_n_arg(command.get(2).map(String::as_str))
-                    && command.get(3).map(String::is_empty) == Some(false)
             } =>
         {
             true

codex-rs/core/src/config_loader/macos.rs

@@ -0,0 +1,118 @@
+use std::io;
+use toml::Value as TomlValue;
+
+#[cfg(target_os = "macos")]
+mod native {
+    use super::*;
+    use base64::Engine;
+    use base64::prelude::BASE64_STANDARD;
+    use core_foundation::base::TCFType;
+    use core_foundation::string::CFString;
+    use core_foundation::string::CFStringRef;
+    use std::ffi::c_void;
+    use tokio::task;
+
+    pub(crate) async fn load_managed_admin_config_layer(
+        override_base64: Option<&str>,
+    ) -> io::Result<Option<TomlValue>> {
+        if let Some(encoded) = override_base64 {
+            let trimmed = encoded.trim();
+            return if trimmed.is_empty() {
+                Ok(None)
+            } else {
+                parse_managed_preferences_base64(trimmed).map(Some)
+            };
+        }
+
+        const LOAD_ERROR: &str = "Failed to load managed preferences configuration";
+
+        match task::spawn_blocking(load_managed_admin_config).await {
+            Ok(result) => result,
+            Err(join_err) => {
+                if join_err.is_cancelled() {
+                    tracing::error!("Managed preferences load task was cancelled");
+                } else {
+                    tracing::error!("Managed preferences load task failed: {join_err}");
+                }
+                Err(io::Error::other(LOAD_ERROR))
+            }
+        }
+    }
+
+    pub(super) fn load_managed_admin_config() -> io::Result<Option<TomlValue>> {
+        #[link(name = "CoreFoundation", kind = "framework")]
+        unsafe extern "C" {
+            fn CFPreferencesCopyAppValue(
+                key: CFStringRef,
+                application_id: CFStringRef,
+            ) -> *mut c_void;
+        }
+
+        const MANAGED_PREFERENCES_APPLICATION_ID: &str = "com.openai.codex";
+        const MANAGED_PREFERENCES_CONFIG_KEY: &str = "config_toml_base64";
+
+        let application_id = CFString::new(MANAGED_PREFERENCES_APPLICATION_ID);
+        let key = CFString::new(MANAGED_PREFERENCES_CONFIG_KEY);
+
+        let value_ref = unsafe {
+            CFPreferencesCopyAppValue(
+                key.as_concrete_TypeRef(),
+                application_id.as_concrete_TypeRef(),
+            )
+        };
+
+        if value_ref.is_null() {
+            tracing::debug!(
+                "Managed preferences for {} key {} not found",
+                MANAGED_PREFERENCES_APPLICATION_ID,
+                MANAGED_PREFERENCES_CONFIG_KEY
+            );
+            return Ok(None);
+        }
+
+        let value = unsafe { CFString::wrap_under_create_rule(value_ref as _) };
+        let contents = value.to_string();
+        let trimmed = contents.trim();
+
+        parse_managed_preferences_base64(trimmed).map(Some)
+    }
+
+    pub(super) fn parse_managed_preferences_base64(encoded: &str) -> io::Result<TomlValue> {
+        let decoded = BASE64_STANDARD.decode(encoded.as_bytes()).map_err(|err| {
+            tracing::error!("Failed to decode managed preferences as base64: {err}");
+            io::Error::new(io::ErrorKind::InvalidData, err)
+        })?;
+
+        let decoded_str = String::from_utf8(decoded).map_err(|err| {
+            tracing::error!("Managed preferences base64 contents were not valid UTF-8: {err}");
+            io::Error::new(io::ErrorKind::InvalidData, err)
+        })?;
+
+        match toml::from_str::<TomlValue>(&decoded_str) {
+            Ok(TomlValue::Table(parsed)) => Ok(TomlValue::Table(parsed)),
+            Ok(other) => {
+                tracing::error!(
+                    "Managed preferences TOML must have a table at the root, found {other:?}",
+                );
+                Err(io::Error::new(
+                    io::ErrorKind::InvalidData,
+                    "managed preferences root must be a table",
+                ))
+            }
+            Err(err) => {
+                tracing::error!("Failed to parse managed preferences TOML: {err}");
+                Err(io::Error::new(io::ErrorKind::InvalidData, err))
+            }
+        }
+    }
+}
+
+#[cfg(target_os = "macos")]
+pub(crate) use native::load_managed_admin_config_layer;
+
+#[cfg(not(target_os = "macos"))]
+pub(crate) async fn load_managed_admin_config_layer(
+    _override_base64: Option<&str>,
+) -> io::Result<Option<TomlValue>> {
+    Ok(None)
+}

codex-rs/core/src/config_loader/mod.rs

@@ -0,0 +1,311 @@
+mod macos;
+
+use crate::config::CONFIG_TOML_FILE;
+use macos::load_managed_admin_config_layer;
+use std::io;
+use std::path::Path;
+use std::path::PathBuf;
+use tokio::fs;
+use toml::Value as TomlValue;
+
+#[cfg(unix)]
+const CODEX_MANAGED_CONFIG_SYSTEM_PATH: &str = "/etc/codex/managed_config.toml";
+
+#[derive(Debug)]
+pub(crate) struct LoadedConfigLayers {
+    pub base: TomlValue,
+    pub managed_config: Option<TomlValue>,
+    pub managed_preferences: Option<TomlValue>,
+}
+
+#[derive(Debug, Default)]
+pub(crate) struct LoaderOverrides {
+    pub managed_config_path: Option<PathBuf>,
+    #[cfg(target_os = "macos")]
+    pub managed_preferences_base64: Option<String>,
+}
+
+// Configuration layering pipeline (top overrides bottom):
+//
+//        +-------------------------+
+//        | Managed preferences (*) |
+//        +-------------------------+
+//                    ^
+//                    |
+//        +-------------------------+
+//        |  managed_config.toml   |
+//        +-------------------------+
+//                    ^
+//                    |
+//        +-------------------------+
+//        |    config.toml (base)   |
+//        +-------------------------+
+//
+// (*) Only available on macOS via managed device profiles.
+
+pub async fn load_config_as_toml(codex_home: &Path) -> io::Result<TomlValue> {
+    load_config_as_toml_with_overrides(codex_home, LoaderOverrides::default()).await
+}
+
+fn default_empty_table() -> TomlValue {
+    TomlValue::Table(Default::default())
+}
+
+pub(crate) async fn load_config_layers_with_overrides(
+    codex_home: &Path,
+    overrides: LoaderOverrides,
+) -> io::Result<LoadedConfigLayers> {
+    load_config_layers_internal(codex_home, overrides).await
+}
+
+async fn load_config_as_toml_with_overrides(
+    codex_home: &Path,
+    overrides: LoaderOverrides,
+) -> io::Result<TomlValue> {
+    let layers = load_config_layers_internal(codex_home, overrides).await?;
+    Ok(apply_managed_layers(layers))
+}
+
+async fn load_config_layers_internal(
+    codex_home: &Path,
+    overrides: LoaderOverrides,
+) -> io::Result<LoadedConfigLayers> {
+    #[cfg(target_os = "macos")]
+    let LoaderOverrides {
+        managed_config_path,
+        managed_preferences_base64,
+    } = overrides;
+
+    #[cfg(not(target_os = "macos"))]
+    let LoaderOverrides {
+        managed_config_path,
+    } = overrides;
+
+    let managed_config_path =
+        managed_config_path.unwrap_or_else(|| managed_config_default_path(codex_home));
+
+    let user_config_path = codex_home.join(CONFIG_TOML_FILE);
+    let user_config = read_config_from_path(&user_config_path, true).await?;
+    let managed_config = read_config_from_path(&managed_config_path, false).await?;
+
+    #[cfg(target_os = "macos")]
+    let managed_preferences =
+        load_managed_admin_config_layer(managed_preferences_base64.as_deref()).await?;
+
+    #[cfg(not(target_os = "macos"))]
+    let managed_preferences = load_managed_admin_config_layer(None).await?;
+
+    Ok(LoadedConfigLayers {
+        base: user_config.unwrap_or_else(default_empty_table),
+        managed_config,
+        managed_preferences,
+    })
+}
+
+async fn read_config_from_path(
+    path: &Path,
+    log_missing_as_info: bool,
+) -> io::Result<Option<TomlValue>> {
+    match fs::read_to_string(path).await {
+        Ok(contents) => match toml::from_str::<TomlValue>(&contents) {
+            Ok(value) => Ok(Some(value)),
+            Err(err) => {
+                tracing::error!("Failed to parse {}: {err}", path.display());
+                Err(io::Error::new(io::ErrorKind::InvalidData, err))
+            }
+        },
+        Err(err) if err.kind() == io::ErrorKind::NotFound => {
+            if log_missing_as_info {
+                tracing::info!("{} not found, using defaults", path.display());
+            } else {
+                tracing::debug!("{} not found", path.display());
+            }
+            Ok(None)
+        }
+        Err(err) => {
+            tracing::error!("Failed to read {}: {err}", path.display());
+            Err(err)
+        }
+    }
+}
+
+/// Merge config `overlay` into `base`, giving `overlay` precedence.
+pub(crate) fn merge_toml_values(base: &mut TomlValue, overlay: &TomlValue) {
+    if let TomlValue::Table(overlay_table) = overlay
+        && let TomlValue::Table(base_table) = base
+    {
+        for (key, value) in overlay_table {
+            if let Some(existing) = base_table.get_mut(key) {
+                merge_toml_values(existing, value);
+            } else {
+                base_table.insert(key.clone(), value.clone());
+            }
+        }
+    } else {
+        *base = overlay.clone();
+    }
+}
+
+fn managed_config_default_path(codex_home: &Path) -> PathBuf {
+    #[cfg(unix)]
+    {
+        let _ = codex_home;
+        PathBuf::from(CODEX_MANAGED_CONFIG_SYSTEM_PATH)
+    }
+
+    #[cfg(not(unix))]
+    {
+        codex_home.join("managed_config.toml")
+    }
+}
+
+fn apply_managed_layers(layers: LoadedConfigLayers) -> TomlValue {
+    let LoadedConfigLayers {
+        mut base,
+        managed_config,
+        managed_preferences,
+    } = layers;
+
+    for overlay in [managed_config, managed_preferences].into_iter().flatten() {
+        merge_toml_values(&mut base, &overlay);
+    }
+
+    base
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use tempfile::tempdir;
+
+    #[tokio::test]
+    async fn merges_managed_config_layer_on_top() {
+        let tmp = tempdir().expect("tempdir");
+        let managed_path = tmp.path().join("managed_config.toml");
+
+        std::fs::write(
+            tmp.path().join(CONFIG_TOML_FILE),
+            r#"foo = 1
+
+[nested]
+value = "base"
+"#,
+        )
+        .expect("write base");
+        std::fs::write(
+            &managed_path,
+            r#"foo = 2
+
+[nested]
+value = "managed_config"
+extra = true
+"#,
+        )
+        .expect("write managed config");
+
+        let overrides = LoaderOverrides {
+            managed_config_path: Some(managed_path),
+            #[cfg(target_os = "macos")]
+            managed_preferences_base64: None,
+        };
+
+        let loaded = load_config_as_toml_with_overrides(tmp.path(), overrides)
+            .await
+            .expect("load config");
+        let table = loaded.as_table().expect("top-level table expected");
+
+        assert_eq!(table.get("foo"), Some(&TomlValue::Integer(2)));
+        let nested = table
+            .get("nested")
+            .and_then(|v| v.as_table())
+            .expect("nested");
+        assert_eq!(
+            nested.get("value"),
+            Some(&TomlValue::String("managed_config".to_string()))
+        );
+        assert_eq!(nested.get("extra"), Some(&TomlValue::Boolean(true)));
+    }
+
+    #[tokio::test]
+    async fn returns_empty_when_all_layers_missing() {
+        let tmp = tempdir().expect("tempdir");
+        let managed_path = tmp.path().join("managed_config.toml");
+        let overrides = LoaderOverrides {
+            managed_config_path: Some(managed_path),
+            #[cfg(target_os = "macos")]
+            managed_preferences_base64: None,
+        };
+
+        let layers = load_config_layers_with_overrides(tmp.path(), overrides)
+            .await
+            .expect("load layers");
+        let base_table = layers.base.as_table().expect("base table expected");
+        assert!(
+            base_table.is_empty(),
+            "expected empty base layer when configs missing"
+        );
+        assert!(
+            layers.managed_config.is_none(),
+            "managed config layer should be absent when file missing"
+        );
+
+        #[cfg(not(target_os = "macos"))]
+        {
+            let loaded = load_config_as_toml(tmp.path()).await.expect("load config");
+            let table = loaded.as_table().expect("top-level table expected");
+            assert!(
+                table.is_empty(),
+                "expected empty table when configs missing"
+            );
+        }
+    }
+
+    #[cfg(target_os = "macos")]
+    #[tokio::test]
+    async fn managed_preferences_take_highest_precedence() {
+        use base64::Engine;
+
+        let managed_payload = r#"
+[nested]
+value = "managed"
+flag = false
+"#;
+        let encoded = base64::prelude::BASE64_STANDARD.encode(managed_payload.as_bytes());
+        let tmp = tempdir().expect("tempdir");
+        let managed_path = tmp.path().join("managed_config.toml");
+
+        std::fs::write(
+            tmp.path().join(CONFIG_TOML_FILE),
+            r#"[nested]
+value = "base"
+"#,
+        )
+        .expect("write base");
+        std::fs::write(
+            &managed_path,
+            r#"[nested]
+value = "managed_config"
+flag = true
+"#,
+        )
+        .expect("write managed config");
+
+        let overrides = LoaderOverrides {
+            managed_config_path: Some(managed_path),
+            managed_preferences_base64: Some(encoded),
+        };
+
+        let loaded = load_config_as_toml_with_overrides(tmp.path(), overrides)
+            .await
+            .expect("load config");
+        let nested = loaded
+            .get("nested")
+            .and_then(|v| v.as_table())
+            .expect("nested table");
+        assert_eq!(
+            nested.get("value"),
+            Some(&TomlValue::String("managed".to_string()))
+        );
+        assert_eq!(nested.get("flag"), Some(&TomlValue::Boolean(false)));
+    }
+}

codex-rs/core/src/config_profile.rs

@@ -20,6 +20,18 @@ pub struct ConfigProfile {
     pub model_verbosity: Option<Verbosity>,
     pub chatgpt_base_url: Option<String>,
     pub experimental_instructions_file: Option<PathBuf>,
+    pub include_plan_tool: Option<bool>,
+    pub include_apply_patch_tool: Option<bool>,
+    pub include_view_image_tool: Option<bool>,
+    pub experimental_use_unified_exec_tool: Option<bool>,
+    pub experimental_use_exec_command_tool: Option<bool>,
+    pub experimental_use_rmcp_client: Option<bool>,
+    pub experimental_use_freeform_apply_patch: Option<bool>,
+    pub tools_web_search: Option<bool>,
+    pub tools_view_image: Option<bool>,
+    /// Optional feature toggles scoped to this profile.
+    #[serde(default)]
+    pub features: Option<crate::features::FeaturesToml>,
 }
 
 impl From<ConfigProfile> for codex_app_server_protocol::Profile {

codex-rs/core/src/config_types.rs

@@ -20,6 +20,10 @@ pub struct McpServerConfig {
     #[serde(flatten)]
     pub transport: McpServerTransportConfig,
 
+    /// When `false`, Codex skips initializing this MCP server.
+    #[serde(default = "default_enabled")]
+    pub enabled: bool,
+
     /// Startup timeout in seconds for initializing MCP server & initially listing tools.
     #[serde(
         default,
@@ -40,21 +44,34 @@ impl<'de> Deserialize<'de> for McpServerConfig {
     {
         #[derive(Deserialize)]
         struct RawMcpServerConfig {
+            // stdio
             command: Option<String>,
             #[serde(default)]
             args: Option<Vec<String>>,
             #[serde(default)]
             env: Option<HashMap<String, String>>,
+            #[serde(default)]
+            env_vars: Option<Vec<String>>,
+            #[serde(default)]
+            cwd: Option<PathBuf>,
+            http_headers: Option<HashMap<String, String>>,
+            #[serde(default)]
+            env_http_headers: Option<HashMap<String, String>>,
 
+            // streamable_http
             url: Option<String>,
             bearer_token: Option<String>,
+            bearer_token_env_var: Option<String>,
 
+            // shared
             #[serde(default)]
             startup_timeout_sec: Option<f64>,
             #[serde(default)]
             startup_timeout_ms: Option<u64>,
             #[serde(default, with = "option_duration_secs")]
             tool_timeout_sec: Option<Duration>,
+            #[serde(default)]
+            enabled: Option<bool>,
         }
 
         let raw = RawMcpServerConfig::deserialize(deserializer)?;
@@ -85,30 +102,58 @@ impl<'de> Deserialize<'de> for McpServerConfig {
                 command: Some(command),
                 args,
                 env,
+                env_vars,
+                cwd,
                 url,
-                bearer_token,
+                bearer_token_env_var,
+                http_headers,
+                env_http_headers,
                 ..
             } => {
                 throw_if_set("stdio", "url", url.as_ref())?;
-                throw_if_set("stdio", "bearer_token", bearer_token.as_ref())?;
+                throw_if_set(
+                    "stdio",
+                    "bearer_token_env_var",
+                    bearer_token_env_var.as_ref(),
+                )?;
+                throw_if_set("stdio", "http_headers", http_headers.as_ref())?;
+                throw_if_set("stdio", "env_http_headers", env_http_headers.as_ref())?;
                 McpServerTransportConfig::Stdio {
                     command,
                     args: args.unwrap_or_default(),
                     env,
+                    env_vars: env_vars.unwrap_or_default(),
+                    cwd,
                 }
             }
             RawMcpServerConfig {
                 url: Some(url),
                 bearer_token,
+                bearer_token_env_var,
                 command,
                 args,
                 env,
-                ..
+                env_vars,
+                cwd,
+                http_headers,
+                env_http_headers,
+                startup_timeout_sec: _,
+                tool_timeout_sec: _,
+                startup_timeout_ms: _,
+                enabled: _,
             } => {
                 throw_if_set("streamable_http", "command", command.as_ref())?;
                 throw_if_set("streamable_http", "args", args.as_ref())?;
                 throw_if_set("streamable_http", "env", env.as_ref())?;
-                McpServerTransportConfig::StreamableHttp { url, bearer_token }
+                throw_if_set("streamable_http", "env_vars", env_vars.as_ref())?;
+                throw_if_set("streamable_http", "cwd", cwd.as_ref())?;
+                throw_if_set("streamable_http", "bearer_token", bearer_token.as_ref())?;
+                McpServerTransportConfig::StreamableHttp {
+                    url,
+                    bearer_token_env_var,
+                    http_headers,
+                    env_http_headers,
+                }
             }
             _ => return Err(SerdeError::custom("invalid transport")),
         };
@@ -117,10 +162,15 @@ impl<'de> Deserialize<'de> for McpServerConfig {
             transport,
             startup_timeout_sec,
             tool_timeout_sec: raw.tool_timeout_sec,
+            enabled: raw.enabled.unwrap_or_else(default_enabled),
         })
     }
 }
 
+const fn default_enabled() -> bool {
+    true
+}
+
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq)]
 #[serde(untagged, deny_unknown_fields, rename_all = "snake_case")]
 pub enum McpServerTransportConfig {
@@ -131,15 +181,25 @@ pub enum McpServerTransportConfig {
         args: Vec<String>,
         #[serde(default, skip_serializing_if = "Option::is_none")]
         env: Option<HashMap<String, String>>,
+        #[serde(default, skip_serializing_if = "Vec::is_empty")]
+        env_vars: Vec<String>,
+        #[serde(default, skip_serializing_if = "Option::is_none")]
+        cwd: Option<PathBuf>,
     },
     /// https://modelcontextprotocol.io/specification/2025-06-18/basic/transports#streamable-http
     StreamableHttp {
         url: String,
-        /// A plain text bearer token to use for authentication.
-        /// This bearer token will be included in the HTTP request header as an `Authorization: Bearer <token>` header.
-        /// This should be used with caution because it lives on disk in clear text.
+        /// Name of the environment variable to read for an HTTP bearer token.
+        /// When set, requests will include the token via `Authorization: Bearer <token>`.
+        /// The actual secret value must be provided via the environment.
         #[serde(default, skip_serializing_if = "Option::is_none")]
-        bearer_token: Option<String>,
+        bearer_token_env_var: Option<String>,
+        /// Additional HTTP headers to include in requests to this server.
+        #[serde(default, skip_serializing_if = "Option::is_none")]
+        http_headers: Option<HashMap<String, String>>,
+        /// HTTP headers where the value is sourced from an environment variable.
+        #[serde(default, skip_serializing_if = "Option::is_none")]
+        env_http_headers: Option<HashMap<String, String>>,
     },
 }
 
@@ -301,6 +361,20 @@ pub struct Tui {
     pub notifications: Notifications,
 }
 
+/// Settings for notices we display to users via the tui and app-server clients
+/// (primarily the Codex IDE extension). NOTE: these are different from
+/// notifications - notices are warnings, NUX screens, acknowledgements, etc.
+#[derive(Deserialize, Debug, Clone, PartialEq, Default)]
+pub struct Notice {
+    /// Tracks whether the user has acknowledged the full access warning prompt.
+    pub hide_full_access_warning: Option<bool>,
+}
+
+impl Notice {
+    /// used by set_hide_full_access_warning until we refactor config updates
+    pub(crate) const TABLE_KEY: &'static str = "notice";
+}
+
 #[derive(Deserialize, Debug, Clone, PartialEq, Default)]
 pub struct SandboxWorkspaceWrite {
     #[serde(default)]
@@ -447,9 +521,12 @@ mod tests {
             McpServerTransportConfig::Stdio {
                 command: "echo".to_string(),
                 args: vec![],
-                env: None
+                env: None,
+                env_vars: Vec::new(),
+                cwd: None,
             }
         );
+        assert!(cfg.enabled);
     }
 
     #[test]
@@ -467,9 +544,12 @@ mod tests {
             McpServerTransportConfig::Stdio {
                 command: "echo".to_string(),
                 args: vec!["hello".to_string(), "world".to_string()],
-                env: None
+                env: None,
+                env_vars: Vec::new(),
+                cwd: None,
             }
         );
+        assert!(cfg.enabled);
     }
 
     #[test]
@@ -488,9 +568,69 @@ mod tests {
             McpServerTransportConfig::Stdio {
                 command: "echo".to_string(),
                 args: vec!["hello".to_string(), "world".to_string()],
-                env: Some(HashMap::from([("FOO".to_string(), "BAR".to_string())]))
+                env: Some(HashMap::from([("FOO".to_string(), "BAR".to_string())])),
+                env_vars: Vec::new(),
+                cwd: None,
             }
         );
+        assert!(cfg.enabled);
+    }
+
+    #[test]
+    fn deserialize_stdio_command_server_config_with_env_vars() {
+        let cfg: McpServerConfig = toml::from_str(
+            r#"
+            command = "echo"
+            env_vars = ["FOO", "BAR"]
+        "#,
+        )
+        .expect("should deserialize command config with env_vars");
+
+        assert_eq!(
+            cfg.transport,
+            McpServerTransportConfig::Stdio {
+                command: "echo".to_string(),
+                args: vec![],
+                env: None,
+                env_vars: vec!["FOO".to_string(), "BAR".to_string()],
+                cwd: None,
+            }
+        );
+    }
+
+    #[test]
+    fn deserialize_stdio_command_server_config_with_cwd() {
+        let cfg: McpServerConfig = toml::from_str(
+            r#"
+            command = "echo"
+            cwd = "/tmp"
+        "#,
+        )
+        .expect("should deserialize command config with cwd");
+
+        assert_eq!(
+            cfg.transport,
+            McpServerTransportConfig::Stdio {
+                command: "echo".to_string(),
+                args: vec![],
+                env: None,
+                env_vars: Vec::new(),
+                cwd: Some(PathBuf::from("/tmp")),
+            }
+        );
+    }
+
+    #[test]
+    fn deserialize_disabled_server_config() {
+        let cfg: McpServerConfig = toml::from_str(
+            r#"
+            command = "echo"
+            enabled = false
+        "#,
+        )
+        .expect("should deserialize disabled server config");
+
+        assert!(!cfg.enabled);
     }
 
     #[test]
@@ -506,17 +646,20 @@ mod tests {
             cfg.transport,
             McpServerTransportConfig::StreamableHttp {
                 url: "https://example.com/mcp".to_string(),
-                bearer_token: None
+                bearer_token_env_var: None,
+                http_headers: None,
+                env_http_headers: None,
             }
         );
+        assert!(cfg.enabled);
     }
 
     #[test]
-    fn deserialize_streamable_http_server_config_with_bearer_token() {
+    fn deserialize_streamable_http_server_config_with_env_var() {
         let cfg: McpServerConfig = toml::from_str(
             r#"
             url = "https://example.com/mcp"
-            bearer_token = "secret"
+            bearer_token_env_var = "GITHUB_TOKEN"
         "#,
         )
         .expect("should deserialize http config");
@@ -525,7 +668,35 @@ mod tests {
             cfg.transport,
             McpServerTransportConfig::StreamableHttp {
                 url: "https://example.com/mcp".to_string(),
-                bearer_token: Some("secret".to_string())
+                bearer_token_env_var: Some("GITHUB_TOKEN".to_string()),
+                http_headers: None,
+                env_http_headers: None,
+            }
+        );
+        assert!(cfg.enabled);
+    }
+
+    #[test]
+    fn deserialize_streamable_http_server_config_with_headers() {
+        let cfg: McpServerConfig = toml::from_str(
+            r#"
+            url = "https://example.com/mcp"
+            http_headers = { "X-Foo" = "bar" }
+            env_http_headers = { "X-Token" = "TOKEN_ENV" }
+        "#,
+        )
+        .expect("should deserialize http config with headers");
+
+        assert_eq!(
+            cfg.transport,
+            McpServerTransportConfig::StreamableHttp {
+                url: "https://example.com/mcp".to_string(),
+                bearer_token_env_var: None,
+                http_headers: Some(HashMap::from([("X-Foo".to_string(), "bar".to_string())])),
+                env_http_headers: Some(HashMap::from([(
+                    "X-Token".to_string(),
+                    "TOKEN_ENV".to_string()
+                )])),
             }
         );
     }
@@ -553,13 +724,37 @@ mod tests {
     }
 
     #[test]
-    fn deserialize_rejects_bearer_token_for_stdio_transport() {
+    fn deserialize_rejects_headers_for_stdio() {
         toml::from_str::<McpServerConfig>(
             r#"
             command = "echo"
+            http_headers = { "X-Foo" = "bar" }
+        "#,
+        )
+        .expect_err("should reject http_headers for stdio transport");
+
+        toml::from_str::<McpServerConfig>(
+            r#"
+            command = "echo"
+            env_http_headers = { "X-Foo" = "BAR_ENV" }
+        "#,
+        )
+        .expect_err("should reject env_http_headers for stdio transport");
+    }
+
+    #[test]
+    fn deserialize_rejects_inline_bearer_token_field() {
+        let err = toml::from_str::<McpServerConfig>(
+            r#"
+            url = "https://example.com"
             bearer_token = "secret"
         "#,
         )
-        .expect_err("should reject bearer token for stdio transport");
+        .expect_err("should reject bearer_token field");
+
+        assert!(
+            err.to_string().contains("bearer_token is not supported"),
+            "unexpected error: {err}"
+        );
     }
 }

codex-rs/core/src/conversation_manager.rs

@@ -17,6 +17,7 @@ use codex_protocol::ConversationId;
 use codex_protocol::models::ResponseItem;
 use codex_protocol::protocol::InitialHistory;
 use codex_protocol::protocol::RolloutItem;
+use codex_protocol::protocol::SessionSource;
 use std::collections::HashMap;
 use std::path::PathBuf;
 use std::sync::Arc;
@@ -35,20 +36,25 @@ pub struct NewConversation {
 pub struct ConversationManager {
     conversations: Arc<RwLock<HashMap<ConversationId, Arc<CodexConversation>>>>,
     auth_manager: Arc<AuthManager>,
+    session_source: SessionSource,
 }
 
 impl ConversationManager {
-    pub fn new(auth_manager: Arc<AuthManager>) -> Self {
+    pub fn new(auth_manager: Arc<AuthManager>, session_source: SessionSource) -> Self {
         Self {
             conversations: Arc::new(RwLock::new(HashMap::new())),
             auth_manager,
+            session_source,
         }
     }
 
     /// Construct with a dummy AuthManager containing the provided CodexAuth.
     /// Used for integration tests: should not be used by ordinary business logic.
     pub fn with_auth(auth: CodexAuth) -> Self {
-        Self::new(crate::AuthManager::from_auth_for_testing(auth))
+        Self::new(
+            crate::AuthManager::from_auth_for_testing(auth),
+            SessionSource::Exec,
+        )
     }
 
     pub async fn new_conversation(&self, config: Config) -> CodexResult<NewConversation> {
@@ -64,7 +70,13 @@ impl ConversationManager {
         let CodexSpawnOk {
             codex,
             conversation_id,
-        } = Codex::spawn(config, auth_manager, InitialHistory::New).await?;
+        } = Codex::spawn(
+            config,
+            auth_manager,
+            InitialHistory::New,
+            self.session_source,
+        )
+        .await?;
         self.finalize_spawn(codex, conversation_id).await
     }
 
@@ -121,7 +133,7 @@ impl ConversationManager {
         let CodexSpawnOk {
             codex,
             conversation_id,
-        } = Codex::spawn(config, auth_manager, initial_history).await?;
+        } = Codex::spawn(config, auth_manager, initial_history, self.session_source).await?;
         self.finalize_spawn(codex, conversation_id).await
     }
 
@@ -155,7 +167,7 @@ impl ConversationManager {
         let CodexSpawnOk {
             codex,
             conversation_id,
-        } = Codex::spawn(config, auth_manager, history).await?;
+        } = Codex::spawn(config, auth_manager, history, self.session_source).await?;
 
         self.finalize_spawn(codex, conversation_id).await
     }
@@ -198,6 +210,7 @@ fn truncate_before_nth_user_message(history: InitialHistory, n: usize) -> Initia
 mod tests {
     use super::*;
     use crate::codex::make_session_and_context;
+    use assert_matches::assert_matches;
     use codex_protocol::models::ContentItem;
     use codex_protocol::models::ReasoningItemReasoningSummary;
     use codex_protocol::models::ResponseItem;
@@ -224,7 +237,7 @@ mod tests {
 
     #[test]
     fn drops_from_last_user_only() {
-        let items = vec![
+        let items = [
             user_msg("u1"),
             assistant_msg("a1"),
             assistant_msg("a2"),
@@ -271,7 +284,7 @@ mod tests {
             .map(RolloutItem::ResponseItem)
             .collect();
         let truncated2 = truncate_before_nth_user_message(InitialHistory::Forked(initial2), 2);
-        assert!(matches!(truncated2, InitialHistory::New));
+        assert_matches!(truncated2, InitialHistory::New);
     }
 
     #[test]

codex-rs/core/src/default_client.rs

@@ -20,7 +20,7 @@ use std::sync::OnceLock;
 /// The full user agent string is returned from the mcp initialize response.
 /// Parenthesis will be added by Codex. This should only specify what goes inside of the parenthesis.
 pub static USER_AGENT_SUFFIX: LazyLock<Mutex<Option<String>>> = LazyLock::new(|| Mutex::new(None));
-
+pub const DEFAULT_ORIGINATOR: &str = "codex_cli_rs";
 pub const CODEX_INTERNAL_ORIGINATOR_OVERRIDE_ENV_VAR: &str = "CODEX_INTERNAL_ORIGINATOR_OVERRIDE";
 #[derive(Debug, Clone)]
 pub struct Originator {
@@ -35,10 +35,11 @@ pub enum SetOriginatorError {
     AlreadyInitialized,
 }
 
-fn init_originator_from_env() -> Originator {
-    let default = "codex_cli_rs";
+fn get_originator_value(provided: Option<String>) -> Originator {
     let value = std::env::var(CODEX_INTERNAL_ORIGINATOR_OVERRIDE_ENV_VAR)
-        .unwrap_or_else(|_| default.to_string());
+        .ok()
+        .or(provided)
+        .unwrap_or(DEFAULT_ORIGINATOR.to_string());
 
     match HeaderValue::from_str(&value) {
         Ok(header_value) => Originator {
@@ -48,31 +49,22 @@ fn init_originator_from_env() -> Originator {
         Err(e) => {
             tracing::error!("Unable to turn originator override {value} into header value: {e}");
             Originator {
-                value: default.to_string(),
-                header_value: HeaderValue::from_static(default),
+                value: DEFAULT_ORIGINATOR.to_string(),
+                header_value: HeaderValue::from_static(DEFAULT_ORIGINATOR),
             }
         }
     }
 }
 
-fn build_originator(value: String) -> Result<Originator, SetOriginatorError> {
-    let header_value =
-        HeaderValue::from_str(&value).map_err(|_| SetOriginatorError::InvalidHeaderValue)?;
-    Ok(Originator {
-        value,
-        header_value,
-    })
-}
-
-pub fn set_default_originator(value: &str) -> Result<(), SetOriginatorError> {
-    let originator = build_originator(value.to_string())?;
+pub fn set_default_originator(value: String) -> Result<(), SetOriginatorError> {
+    let originator = get_originator_value(Some(value));
     ORIGINATOR
         .set(originator)
         .map_err(|_| SetOriginatorError::AlreadyInitialized)
 }
 
 pub fn originator() -> &'static Originator {
-    ORIGINATOR.get_or_init(init_originator_from_env)
+    ORIGINATOR.get_or_init(|| get_originator_value(None))
 }
 
 pub fn get_codex_user_agent() -> String {

codex-rs/core/src/environment_context.rs

@@ -93,6 +93,25 @@ impl EnvironmentContext {
             && self.network_access == *network_access
             && self.writable_roots == *writable_roots
     }
+
+    pub fn diff(before: &TurnContext, after: &TurnContext) -> Self {
+        let cwd = if before.cwd != after.cwd {
+            Some(after.cwd.clone())
+        } else {
+            None
+        };
+        let approval_policy = if before.approval_policy != after.approval_policy {
+            Some(after.approval_policy)
+        } else {
+            None
+        };
+        let sandbox_policy = if before.sandbox_policy != after.sandbox_policy {
+            Some(after.sandbox_policy.clone())
+        } else {
+            None
+        };
+        EnvironmentContext::new(cwd, approval_policy, sandbox_policy, None)
+    }
 }
 
 impl From<&TurnContext> for EnvironmentContext {

codex-rs/core/src/error.rs

@@ -1,6 +1,10 @@
 use crate::exec::ExecToolCallOutput;
 use crate::token_data::KnownPlan;
 use crate::token_data::PlanType;
+use crate::truncate::truncate_middle;
+use chrono::DateTime;
+use chrono::Utc;
+use codex_async_utils::CancelErr;
 use codex_protocol::ConversationId;
 use codex_protocol::protocol::RateLimitSnapshot;
 use reqwest::StatusCode;
@@ -12,6 +16,9 @@ use tokio::task::JoinError;
 
 pub type Result<T> = std::result::Result<T, CodexErr>;
 
+/// Limit UI error messages to a reasonable size while keeping useful context.
+const ERROR_MESSAGE_UI_MAX_BYTES: usize = 2 * 1024; // 4 KiB
+
 #[derive(Error, Debug)]
 pub enum SandboxErr {
     /// Error from sandbox execution
@@ -46,6 +53,9 @@ pub enum SandboxErr {
 
 #[derive(Error, Debug)]
 pub enum CodexErr {
+    #[error("turn aborted")]
+    TurnAborted,
+
     /// Returned by ResponsesClient when the SSE stream disconnects or errors out **after** the HTTP
     /// handshake has succeeded but **before** it finished emitting `response.completed`.
     ///
@@ -55,6 +65,11 @@ pub enum CodexErr {
     #[error("stream disconnected before completion: {0}")]
     Stream(String, Option<Duration>),
 
+    #[error(
+        "Codex ran out of room in the model's context window. Start a new conversation or clear earlier history before retrying."
+    )]
+    ContextWindowExceeded,
+
     #[error("no conversation with id: {0}")]
     ConversationNotFound(ConversationId),
 
@@ -82,6 +97,12 @@ pub enum CodexErr {
     #[error("{0}")]
     UsageLimitReached(UsageLimitReachedError),
 
+    #[error("{0}")]
+    ResponseStreamFailed(ResponseStreamFailed),
+
+    #[error("{0}")]
+    ConnectionFailed(ConnectionFailedError),
+
     #[error(
         "To use Codex with your ChatGPT plan, upgrade to Plus: https://openai.com/chatgpt/pricing."
     )]
@@ -108,15 +129,15 @@ pub enum CodexErr {
     #[error("unsupported operation: {0}")]
     UnsupportedOperation(String),
 
+    #[error("Fatal error: {0}")]
+    Fatal(String),
+
     // -----------------------------------------------------------------
     // Automatic conversions for common external error types
     // -----------------------------------------------------------------
     #[error(transparent)]
     Io(#[from] io::Error),
 
-    #[error(transparent)]
-    Reqwest(#[from] reqwest::Error),
-
     #[error(transparent)]
     Json(#[from] serde_json::Error),
 
@@ -135,6 +156,43 @@ pub enum CodexErr {
     EnvVar(EnvVarError),
 }
 
+impl From<CancelErr> for CodexErr {
+    fn from(_: CancelErr) -> Self {
+        CodexErr::TurnAborted
+    }
+}
+
+#[derive(Debug)]
+pub struct ConnectionFailedError {
+    pub source: reqwest::Error,
+}
+
+impl std::fmt::Display for ConnectionFailedError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "Connection failed: {}", self.source)
+    }
+}
+
+#[derive(Debug)]
+pub struct ResponseStreamFailed {
+    pub source: reqwest::Error,
+    pub request_id: Option<String>,
+}
+
+impl std::fmt::Display for ResponseStreamFailed {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(
+            f,
+            "Error while reading the server response: {}{}",
+            self.source,
+            self.request_id
+                .as_ref()
+                .map(|id| format!(", request id: {id}"))
+                .unwrap_or_default()
+        )
+    }
+}
+
 #[derive(Debug)]
 pub struct UnexpectedResponseError {
     pub status: StatusCode,
@@ -181,7 +239,7 @@ impl std::fmt::Display for RetryLimitReachedError {
 #[derive(Debug)]
 pub struct UsageLimitReachedError {
     pub(crate) plan_type: Option<PlanType>,
-    pub(crate) resets_in_seconds: Option<u64>,
+    pub(crate) resets_at: Option<DateTime<Utc>>,
     pub(crate) rate_limits: Option<RateLimitSnapshot>,
 }
 
@@ -190,12 +248,12 @@ impl std::fmt::Display for UsageLimitReachedError {
         let message = match self.plan_type.as_ref() {
             Some(PlanType::Known(KnownPlan::Plus)) => format!(
                 "You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing){}",
-                retry_suffix_after_or(self.resets_in_seconds)
+                retry_suffix_after_or(self.resets_at.as_ref())
             ),
             Some(PlanType::Known(KnownPlan::Team)) | Some(PlanType::Known(KnownPlan::Business)) => {
                 format!(
                     "You've hit your usage limit. To get more access now, send a request to your admin{}",
-                    retry_suffix_after_or(self.resets_in_seconds)
+                    retry_suffix_after_or(self.resets_at.as_ref())
                 )
             }
             Some(PlanType::Known(KnownPlan::Free)) => {
@@ -206,11 +264,11 @@ impl std::fmt::Display for UsageLimitReachedError {
             | Some(PlanType::Known(KnownPlan::Enterprise))
             | Some(PlanType::Known(KnownPlan::Edu)) => format!(
                 "You've hit your usage limit.{}",
-                retry_suffix(self.resets_in_seconds)
+                retry_suffix(self.resets_at.as_ref())
             ),
             Some(PlanType::Unknown(_)) | None => format!(
                 "You've hit your usage limit.{}",
-                retry_suffix(self.resets_in_seconds)
+                retry_suffix(self.resets_at.as_ref())
             ),
         };
 
@@ -218,8 +276,8 @@ impl std::fmt::Display for UsageLimitReachedError {
     }
 }
 
-fn retry_suffix(resets_in_seconds: Option<u64>) -> String {
-    if let Some(secs) = resets_in_seconds {
+fn retry_suffix(resets_at: Option<&DateTime<Utc>>) -> String {
+    if let Some(secs) = remaining_seconds(resets_at) {
         let reset_duration = format_reset_duration(secs);
         format!(" Try again in {reset_duration}.")
     } else {
@@ -227,8 +285,8 @@ fn retry_suffix(resets_in_seconds: Option<u64>) -> String {
     }
 }
 
-fn retry_suffix_after_or(resets_in_seconds: Option<u64>) -> String {
-    if let Some(secs) = resets_in_seconds {
+fn retry_suffix_after_or(resets_at: Option<&DateTime<Utc>>) -> String {
+    if let Some(secs) = remaining_seconds(resets_at) {
         let reset_duration = format_reset_duration(secs);
         format!(" or try again in {reset_duration}.")
     } else {
@@ -236,6 +294,29 @@ fn retry_suffix_after_or(resets_in_seconds: Option<u64>) -> String {
     }
 }
 
+fn remaining_seconds(resets_at: Option<&DateTime<Utc>>) -> Option<u64> {
+    let resets_at = resets_at.cloned()?;
+    let now = now_for_retry();
+    let secs = resets_at.signed_duration_since(now).num_seconds();
+    Some(if secs <= 0 { 0 } else { secs as u64 })
+}
+
+#[cfg(test)]
+thread_local! {
+    static NOW_OVERRIDE: std::cell::RefCell<Option<DateTime<Utc>>> =
+        const { std::cell::RefCell::new(None) };
+}
+
+fn now_for_retry() -> DateTime<Utc> {
+    #[cfg(test)]
+    {
+        if let Some(now) = NOW_OVERRIDE.with(|cell| *cell.borrow()) {
+            return now;
+        }
+    }
+    Utc::now()
+}
+
 fn format_reset_duration(total_secs: u64) -> String {
     let days = total_secs / 86_400;
     let hours = (total_secs % 86_400) / 3_600;
@@ -296,42 +377,86 @@ impl CodexErr {
 }
 
 pub fn get_error_message_ui(e: &CodexErr) -> String {
-    match e {
-        CodexErr::Sandbox(SandboxErr::Denied { output }) => output.stderr.text.clone(),
+    let message = match e {
+        CodexErr::Sandbox(SandboxErr::Denied { output }) => {
+            let aggregated = output.aggregated_output.text.trim();
+            if !aggregated.is_empty() {
+                output.aggregated_output.text.clone()
+            } else {
+                let stderr = output.stderr.text.trim();
+                let stdout = output.stdout.text.trim();
+                match (stderr.is_empty(), stdout.is_empty()) {
+                    (false, false) => format!("{stderr}\n{stdout}"),
+                    (false, true) => output.stderr.text.clone(),
+                    (true, false) => output.stdout.text.clone(),
+                    (true, true) => format!(
+                        "command failed inside sandbox with exit code {}",
+                        output.exit_code
+                    ),
+                }
+            }
+        }
         // Timeouts are not sandbox errors from a UX perspective; present them plainly
-        CodexErr::Sandbox(SandboxErr::Timeout { output }) => format!(
-            "error: command timed out after {} ms",
-            output.duration.as_millis()
-        ),
+        CodexErr::Sandbox(SandboxErr::Timeout { output }) => {
+            format!(
+                "error: command timed out after {} ms",
+                output.duration.as_millis()
+            )
+        }
         _ => e.to_string(),
-    }
+    };
+
+    truncate_middle(&message, ERROR_MESSAGE_UI_MAX_BYTES).0
 }
 
 #[cfg(test)]
 mod tests {
     use super::*;
+    use crate::exec::StreamOutput;
+    use chrono::DateTime;
+    use chrono::Duration as ChronoDuration;
+    use chrono::TimeZone;
+    use chrono::Utc;
     use codex_protocol::protocol::RateLimitWindow;
+    use pretty_assertions::assert_eq;
 
     fn rate_limit_snapshot() -> RateLimitSnapshot {
+        let primary_reset_at = Utc
+            .with_ymd_and_hms(2024, 1, 1, 1, 0, 0)
+            .unwrap()
+            .timestamp();
+        let secondary_reset_at = Utc
+            .with_ymd_and_hms(2024, 1, 1, 2, 0, 0)
+            .unwrap()
+            .timestamp();
         RateLimitSnapshot {
             primary: Some(RateLimitWindow {
                 used_percent: 50.0,
                 window_minutes: Some(60),
-                resets_in_seconds: Some(3600),
+                resets_at: Some(primary_reset_at),
             }),
             secondary: Some(RateLimitWindow {
                 used_percent: 30.0,
                 window_minutes: Some(120),
-                resets_in_seconds: Some(7200),
+                resets_at: Some(secondary_reset_at),
             }),
         }
     }
 
+    fn with_now_override<T>(now: DateTime<Utc>, f: impl FnOnce() -> T) -> T {
+        NOW_OVERRIDE.with(|cell| {
+            *cell.borrow_mut() = Some(now);
+            let result = f();
+            *cell.borrow_mut() = None;
+            result
+        })
+    }
+
     #[test]
     fn usage_limit_reached_error_formats_plus_plan() {
         let err = UsageLimitReachedError {
             plan_type: Some(PlanType::Known(KnownPlan::Plus)),
-            resets_in_seconds: None,
+            resets_at: None,
             rate_limits: Some(rate_limit_snapshot()),
         };
         assert_eq!(
@@ -340,11 +465,78 @@ mod tests {
         );
     }
 
+    #[test]
+    fn sandbox_denied_uses_aggregated_output_when_stderr_empty() {
+        let output = ExecToolCallOutput {
+            exit_code: 77,
+            stdout: StreamOutput::new(String::new()),
+            stderr: StreamOutput::new(String::new()),
+            aggregated_output: StreamOutput::new("aggregate detail".to_string()),
+            duration: Duration::from_millis(10),
+            timed_out: false,
+        };
+        let err = CodexErr::Sandbox(SandboxErr::Denied {
+            output: Box::new(output),
+        });
+        assert_eq!(get_error_message_ui(&err), "aggregate detail");
+    }
+
+    #[test]
+    fn sandbox_denied_reports_both_streams_when_available() {
+        let output = ExecToolCallOutput {
+            exit_code: 9,
+            stdout: StreamOutput::new("stdout detail".to_string()),
+            stderr: StreamOutput::new("stderr detail".to_string()),
+            aggregated_output: StreamOutput::new(String::new()),
+            duration: Duration::from_millis(10),
+            timed_out: false,
+        };
+        let err = CodexErr::Sandbox(SandboxErr::Denied {
+            output: Box::new(output),
+        });
+        assert_eq!(get_error_message_ui(&err), "stderr detail\nstdout detail");
+    }
+
+    #[test]
+    fn sandbox_denied_reports_stdout_when_no_stderr() {
+        let output = ExecToolCallOutput {
+            exit_code: 11,
+            stdout: StreamOutput::new("stdout only".to_string()),
+            stderr: StreamOutput::new(String::new()),
+            aggregated_output: StreamOutput::new(String::new()),
+            duration: Duration::from_millis(8),
+            timed_out: false,
+        };
+        let err = CodexErr::Sandbox(SandboxErr::Denied {
+            output: Box::new(output),
+        });
+        assert_eq!(get_error_message_ui(&err), "stdout only");
+    }
+
+    #[test]
+    fn sandbox_denied_reports_exit_code_when_no_output_available() {
+        let output = ExecToolCallOutput {
+            exit_code: 13,
+            stdout: StreamOutput::new(String::new()),
+            stderr: StreamOutput::new(String::new()),
+            aggregated_output: StreamOutput::new(String::new()),
+            duration: Duration::from_millis(5),
+            timed_out: false,
+        };
+        let err = CodexErr::Sandbox(SandboxErr::Denied {
+            output: Box::new(output),
+        });
+        assert_eq!(
+            get_error_message_ui(&err),
+            "command failed inside sandbox with exit code 13"
+        );
+    }
+
     #[test]
     fn usage_limit_reached_error_formats_free_plan() {
         let err = UsageLimitReachedError {
             plan_type: Some(PlanType::Known(KnownPlan::Free)),
-            resets_in_seconds: Some(3600),
+            resets_at: None,
             rate_limits: Some(rate_limit_snapshot()),
         };
         assert_eq!(
@@ -357,7 +549,7 @@ mod tests {
     fn usage_limit_reached_error_formats_default_when_none() {
         let err = UsageLimitReachedError {
             plan_type: None,
-            resets_in_seconds: None,
+            resets_at: None,
             rate_limits: Some(rate_limit_snapshot()),
         };
         assert_eq!(
@@ -368,22 +560,26 @@ mod tests {
 
     #[test]
     fn usage_limit_reached_error_formats_team_plan() {
-        let err = UsageLimitReachedError {
-            plan_type: Some(PlanType::Known(KnownPlan::Team)),
-            resets_in_seconds: Some(3600),
-            rate_limits: Some(rate_limit_snapshot()),
-        };
-        assert_eq!(
-            err.to_string(),
-            "You've hit your usage limit. To get more access now, send a request to your admin or try again in 1 hour."
-        );
+        let base = Utc.with_ymd_and_hms(2024, 1, 1, 0, 0, 0).unwrap();
+        let resets_at = base + ChronoDuration::hours(1);
+        with_now_override(base, move || {
+            let err = UsageLimitReachedError {
+                plan_type: Some(PlanType::Known(KnownPlan::Team)),
+                resets_at: Some(resets_at),
+                rate_limits: Some(rate_limit_snapshot()),
+            };
+            assert_eq!(
+                err.to_string(),
+                "You've hit your usage limit. To get more access now, send a request to your admin or try again in 1 hour."
+            );
+        });
     }
 
     #[test]
     fn usage_limit_reached_error_formats_business_plan_without_reset() {
         let err = UsageLimitReachedError {
             plan_type: Some(PlanType::Known(KnownPlan::Business)),
-            resets_in_seconds: None,
+            resets_at: None,
             rate_limits: Some(rate_limit_snapshot()),
         };
         assert_eq!(
@@ -396,7 +592,7 @@ mod tests {
     fn usage_limit_reached_error_formats_default_for_other_plans() {
         let err = UsageLimitReachedError {
             plan_type: Some(PlanType::Known(KnownPlan::Pro)),
-            resets_in_seconds: None,
+            resets_at: None,
             rate_limits: Some(rate_limit_snapshot()),
         };
         assert_eq!(
@@ -407,53 +603,70 @@ mod tests {
 
     #[test]
     fn usage_limit_reached_includes_minutes_when_available() {
-        let err = UsageLimitReachedError {
-            plan_type: None,
-            resets_in_seconds: Some(5 * 60),
-            rate_limits: Some(rate_limit_snapshot()),
-        };
-        assert_eq!(
-            err.to_string(),
-            "You've hit your usage limit. Try again in 5 minutes."
-        );
+        let base = Utc.with_ymd_and_hms(2024, 1, 1, 0, 0, 0).unwrap();
+        let resets_at = base + ChronoDuration::minutes(5);
+        with_now_override(base, move || {
+            let err = UsageLimitReachedError {
+                plan_type: None,
+                resets_at: Some(resets_at),
+                rate_limits: Some(rate_limit_snapshot()),
+            };
+            assert_eq!(
+                err.to_string(),
+                "You've hit your usage limit. Try again in 5 minutes."
+            );
+        });
     }
 
     #[test]
     fn usage_limit_reached_includes_hours_and_minutes() {
-        let err = UsageLimitReachedError {
-            plan_type: Some(PlanType::Known(KnownPlan::Plus)),
-            resets_in_seconds: Some(3 * 3600 + 32 * 60),
-            rate_limits: Some(rate_limit_snapshot()),
-        };
-        assert_eq!(
-            err.to_string(),
-            "You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing) or try again in 3 hours 32 minutes."
-        );
+        let base = Utc.with_ymd_and_hms(2024, 1, 1, 0, 0, 0).unwrap();
+        let resets_at = base + ChronoDuration::hours(3) + ChronoDuration::minutes(32);
+        with_now_override(base, move || {
+            let err = UsageLimitReachedError {
+                plan_type: Some(PlanType::Known(KnownPlan::Plus)),
+                resets_at: Some(resets_at),
+                rate_limits: Some(rate_limit_snapshot()),
+            };
+            assert_eq!(
+                err.to_string(),
+                "You've hit your usage limit. Upgrade to Pro (https://openai.com/chatgpt/pricing) or try again in 3 hours 32 minutes."
+            );
+        });
     }
 
     #[test]
     fn usage_limit_reached_includes_days_hours_minutes() {
-        let err = UsageLimitReachedError {
-            plan_type: None,
-            resets_in_seconds: Some(2 * 86_400 + 3 * 3600 + 5 * 60),
-            rate_limits: Some(rate_limit_snapshot()),
-        };
-        assert_eq!(
-            err.to_string(),
-            "You've hit your usage limit. Try again in 2 days 3 hours 5 minutes."
-        );
+        let base = Utc.with_ymd_and_hms(2024, 1, 1, 0, 0, 0).unwrap();
+        let resets_at =
+            base + ChronoDuration::days(2) + ChronoDuration::hours(3) + ChronoDuration::minutes(5);
+        with_now_override(base, move || {
+            let err = UsageLimitReachedError {
+                plan_type: None,
+                resets_at: Some(resets_at),
+                rate_limits: Some(rate_limit_snapshot()),
+            };
+            assert_eq!(
+                err.to_string(),
+                "You've hit your usage limit. Try again in 2 days 3 hours 5 minutes."
+            );
+        });
     }
 
     #[test]
     fn usage_limit_reached_less_than_minute() {
-        let err = UsageLimitReachedError {
-            plan_type: None,
-            resets_in_seconds: Some(30),
-            rate_limits: Some(rate_limit_snapshot()),
-        };
-        assert_eq!(
-            err.to_string(),
-            "You've hit your usage limit. Try again in less than a minute."
-        );
+        let base = Utc.with_ymd_and_hms(2024, 1, 1, 0, 0, 0).unwrap();
+        let resets_at = base + ChronoDuration::seconds(30);
+        with_now_override(base, move || {
+            let err = UsageLimitReachedError {
+                plan_type: None,
+                resets_at: Some(resets_at),
+                rate_limits: Some(rate_limit_snapshot()),
+            };
+            assert_eq!(
+                err.to_string(),
+                "You've hit your usage limit. Try again in less than a minute."
+            );
+        });
     }
 }

codex-rs/core/src/event_mapping.rs

@@ -127,6 +127,7 @@ mod tests {
     use super::map_response_item_to_event_messages;
     use crate::protocol::EventMsg;
     use crate::protocol::InputMessageKind;
+    use assert_matches::assert_matches;
     use codex_protocol::models::ContentItem;
     use codex_protocol::models::ResponseItem;
     use pretty_assertions::assert_eq;
@@ -158,7 +159,7 @@ mod tests {
         match &events[0] {
             EventMsg::UserMessage(user) => {
                 assert_eq!(user.message, "Hello world");
-                assert!(matches!(user.kind, Some(InputMessageKind::Plain)));
+                assert_matches!(user.kind, Some(InputMessageKind::Plain));
                 assert_eq!(user.images, Some(vec![img1, img2]));
             }
             other => panic!("expected UserMessage, got {other:?}"),

codex-rs/core/src/exec.rs

@@ -18,13 +18,14 @@ use tokio::process::Child;
 use crate::error::CodexErr;
 use crate::error::Result;
 use crate::error::SandboxErr;
-use crate::landlock::spawn_command_under_linux_sandbox;
 use crate::protocol::Event;
 use crate::protocol::EventMsg;
 use crate::protocol::ExecCommandOutputDeltaEvent;
 use crate::protocol::ExecOutputStream;
 use crate::protocol::SandboxPolicy;
-use crate::seatbelt::spawn_command_under_seatbelt;
+use crate::sandboxing::CommandSpec;
+use crate::sandboxing::ExecEnv;
+use crate::sandboxing::SandboxManager;
 use crate::spawn::StdioPolicy;
 use crate::spawn::spawn_child_async;
 
@@ -53,6 +54,7 @@ pub struct ExecParams {
     pub env: HashMap<String, String>,
     pub with_escalated_permissions: Option<bool>,
     pub justification: Option<String>,
+    pub arg0: Option<String>,
 }
 
 impl ExecParams {
@@ -87,57 +89,85 @@ pub async fn process_exec_tool_call(
     codex_linux_sandbox_exe: &Option<PathBuf>,
     stdout_stream: Option<StdoutStream>,
 ) -> Result<ExecToolCallOutput> {
-    let start = Instant::now();
+    let ExecParams {
+        command,
+        cwd,
+        timeout_ms,
+        env,
+        with_escalated_permissions,
+        justification,
+        arg0: _,
+    } = params;
 
-    let timeout_duration = params.timeout_duration();
+    let (program, args) = command.split_first().ok_or_else(|| {
+        CodexErr::Io(io::Error::new(
+            io::ErrorKind::InvalidInput,
+            "command args are empty",
+        ))
+    })?;
 
-    let raw_output_result: std::result::Result<RawExecToolCallOutput, CodexErr> = match sandbox_type
-    {
-        SandboxType::None => exec(params, sandbox_policy, stdout_stream.clone()).await,
-        SandboxType::MacosSeatbelt => {
-            let ExecParams {
-                command,
-                cwd: command_cwd,
-                env,
-                ..
-            } = params;
-            let child = spawn_command_under_seatbelt(
-                command,
-                command_cwd,
-                sandbox_policy,
-                sandbox_cwd,
-                StdioPolicy::RedirectForShellTool,
-                env,
-            )
-            .await?;
-            consume_truncated_output(child, timeout_duration, stdout_stream.clone()).await
-        }
-        SandboxType::LinuxSeccomp => {
-            let ExecParams {
-                command,
-                cwd: command_cwd,
-                env,
-                ..
-            } = params;
-
-            let codex_linux_sandbox_exe = codex_linux_sandbox_exe
-                .as_ref()
-                .ok_or(CodexErr::LandlockSandboxExecutableNotProvided)?;
-            let child = spawn_command_under_linux_sandbox(
-                codex_linux_sandbox_exe,
-                command,
-                command_cwd,
-                sandbox_policy,
-                sandbox_cwd,
-                StdioPolicy::RedirectForShellTool,
-                env,
-            )
-            .await?;
-
-            consume_truncated_output(child, timeout_duration, stdout_stream).await
-        }
+    let spec = CommandSpec {
+        program: program.clone(),
+        args: args.to_vec(),
+        cwd,
+        env,
+        timeout_ms,
+        with_escalated_permissions,
+        justification,
     };
+
+    let manager = SandboxManager::new();
+    let exec_env = manager
+        .transform(
+            &spec,
+            sandbox_policy,
+            sandbox_type,
+            sandbox_cwd,
+            codex_linux_sandbox_exe.as_ref(),
+        )
+        .map_err(CodexErr::from)?;
+
+    // Route through the sandboxing module for a single, unified execution path.
+    crate::sandboxing::execute_env(&exec_env, sandbox_policy, stdout_stream).await
+}
+
+pub(crate) async fn execute_exec_env(
+    env: ExecEnv,
+    sandbox_policy: &SandboxPolicy,
+    stdout_stream: Option<StdoutStream>,
+) -> Result<ExecToolCallOutput> {
+    let ExecEnv {
+        command,
+        cwd,
+        env,
+        timeout_ms,
+        sandbox,
+        with_escalated_permissions,
+        justification,
+        arg0,
+    } = env;
+
+    let params = ExecParams {
+        command,
+        cwd,
+        timeout_ms,
+        env,
+        with_escalated_permissions,
+        justification,
+        arg0,
+    };
+
+    let start = Instant::now();
+    let raw_output_result = exec(params, sandbox_policy, stdout_stream).await;
     let duration = start.elapsed();
+    finalize_exec_result(raw_output_result, sandbox, duration)
+}
+
+fn finalize_exec_result(
+    raw_output_result: std::result::Result<RawExecToolCallOutput, CodexErr>,
+    sandbox_type: SandboxType,
+    duration: Duration,
+) -> Result<ExecToolCallOutput> {
     match raw_output_result {
         Ok(raw_output) => {
             #[allow(unused_mut)]
@@ -177,7 +207,7 @@ pub async fn process_exec_tool_call(
                 }));
             }
 
-            if exit_code != 0 && is_likely_sandbox_denied(sandbox_type, exit_code) {
+            if is_likely_sandbox_denied(sandbox_type, &exec_output) {
                 return Err(CodexErr::Sandbox(SandboxErr::Denied {
                     output: Box::new(exec_output),
                 }));
@@ -192,31 +222,89 @@ pub async fn process_exec_tool_call(
     }
 }
 
+pub(crate) mod errors {
+    use super::CodexErr;
+    use crate::sandboxing::SandboxTransformError;
+
+    impl From<SandboxTransformError> for CodexErr {
+        fn from(err: SandboxTransformError) -> Self {
+            match err {
+                SandboxTransformError::MissingLinuxSandboxExecutable => {
+                    CodexErr::LandlockSandboxExecutableNotProvided
+                }
+            }
+        }
+    }
+}
+
 /// We don't have a fully deterministic way to tell if our command failed
 /// because of the sandbox - a command in the user's zshrc file might hit an
 /// error, but the command itself might fail or succeed for other reasons.
-/// For now, we conservatively check for 'command not found' (exit code 127),
-/// and can add additional cases as necessary.
-fn is_likely_sandbox_denied(sandbox_type: SandboxType, exit_code: i32) -> bool {
-    if sandbox_type == SandboxType::None {
+/// For now, we conservatively check for well known command failure exit codes and
+/// also look for common sandbox denial keywords in the command output.
+pub(crate) fn is_likely_sandbox_denied(
+    sandbox_type: SandboxType,
+    exec_output: &ExecToolCallOutput,
+) -> bool {
+    if sandbox_type == SandboxType::None || exec_output.exit_code == 0 {
         return false;
     }
 
     // Quick rejects: well-known non-sandbox shell exit codes
-    // 127: command not found, 2: misuse of shell builtins
-    if exit_code == 127 {
+    // 2: misuse of shell builtins
+    // 126: permission denied
+    // 127: command not found
+    const SANDBOX_DENIED_KEYWORDS: [&str; 7] = [
+        "operation not permitted",
+        "permission denied",
+        "read-only file system",
+        "seccomp",
+        "sandbox",
+        "landlock",
+        "failed to write file",
+    ];
+
+    let has_sandbox_keyword = [
+        &exec_output.stderr.text,
+        &exec_output.stdout.text,
+        &exec_output.aggregated_output.text,
+    ]
+    .into_iter()
+    .any(|section| {
+        let lower = section.to_lowercase();
+        SANDBOX_DENIED_KEYWORDS
+            .iter()
+            .any(|needle| lower.contains(needle))
+    });
+
+    if has_sandbox_keyword {
+        return true;
+    }
+
+    const QUICK_REJECT_EXIT_CODES: [i32; 3] = [2, 126, 127];
+    if QUICK_REJECT_EXIT_CODES.contains(&exec_output.exit_code) {
         return false;
     }
 
-    // For all other cases, we assume the sandbox is the cause
-    true
+    #[cfg(unix)]
+    {
+        const SIGSYS_CODE: i32 = libc::SIGSYS;
+        if sandbox_type == SandboxType::LinuxSeccomp
+            && exec_output.exit_code == EXIT_CODE_SIGNAL_BASE + SIGSYS_CODE
+        {
+            return true;
+        }
+    }
+
+    false
 }
 
-#[derive(Debug)]
-pub struct StreamOutput<T> {
+#[derive(Debug, Clone)]
+pub struct StreamOutput<T: Clone> {
     pub text: T,
     pub truncated_after_lines: Option<u32>,
 }
+
 #[derive(Debug)]
 struct RawExecToolCallOutput {
     pub exit_status: ExitStatus,
@@ -249,7 +337,7 @@ fn append_all(dst: &mut Vec<u8>, src: &[u8]) {
     dst.extend_from_slice(src);
 }
 
-#[derive(Debug)]
+#[derive(Clone, Debug)]
 pub struct ExecToolCallOutput {
     pub exit_code: i32,
     pub stdout: StreamOutput<String>,
@@ -266,7 +354,11 @@ async fn exec(
 ) -> Result<RawExecToolCallOutput> {
     let timeout = params.timeout_duration();
     let ExecParams {
-        command, cwd, env, ..
+        command,
+        cwd,
+        env,
+        arg0,
+        ..
     } = params;
 
     let (program, args) = command.split_first().ok_or_else(|| {
@@ -275,11 +367,11 @@ async fn exec(
             "command args are empty",
         ))
     })?;
-    let arg0 = None;
+    let arg0_ref = arg0.as_deref();
     let child = spawn_child_async(
         PathBuf::from(program),
         args.into(),
-        arg0,
+        arg0_ref,
         cwd,
         sandbox_policy,
         StdioPolicy::RedirectForShellTool,
@@ -436,3 +528,77 @@ fn synthetic_exit_status(code: i32) -> ExitStatus {
     #[expect(clippy::unwrap_used)]
     std::process::ExitStatus::from_raw(code.try_into().unwrap())
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::time::Duration;
+
+    fn make_exec_output(
+        exit_code: i32,
+        stdout: &str,
+        stderr: &str,
+        aggregated: &str,
+    ) -> ExecToolCallOutput {
+        ExecToolCallOutput {
+            exit_code,
+            stdout: StreamOutput::new(stdout.to_string()),
+            stderr: StreamOutput::new(stderr.to_string()),
+            aggregated_output: StreamOutput::new(aggregated.to_string()),
+            duration: Duration::from_millis(1),
+            timed_out: false,
+        }
+    }
+
+    #[test]
+    fn sandbox_detection_requires_keywords() {
+        let output = make_exec_output(1, "", "", "");
+        assert!(!is_likely_sandbox_denied(
+            SandboxType::LinuxSeccomp,
+            &output
+        ));
+    }
+
+    #[test]
+    fn sandbox_detection_identifies_keyword_in_stderr() {
+        let output = make_exec_output(1, "", "Operation not permitted", "");
+        assert!(is_likely_sandbox_denied(SandboxType::LinuxSeccomp, &output));
+    }
+
+    #[test]
+    fn sandbox_detection_respects_quick_reject_exit_codes() {
+        let output = make_exec_output(127, "", "command not found", "");
+        assert!(!is_likely_sandbox_denied(
+            SandboxType::LinuxSeccomp,
+            &output
+        ));
+    }
+
+    #[test]
+    fn sandbox_detection_ignores_non_sandbox_mode() {
+        let output = make_exec_output(1, "", "Operation not permitted", "");
+        assert!(!is_likely_sandbox_denied(SandboxType::None, &output));
+    }
+
+    #[test]
+    fn sandbox_detection_uses_aggregated_output() {
+        let output = make_exec_output(
+            101,
+            "",
+            "",
+            "cargo failed: Read-only file system when writing target",
+        );
+        assert!(is_likely_sandbox_denied(
+            SandboxType::MacosSeatbelt,
+            &output
+        ));
+    }
+
+    #[cfg(unix)]
+    #[test]
+    fn sandbox_detection_flags_sigsys_exit_code() {
+        let exit_code = EXIT_CODE_SIGNAL_BASE + libc::SIGSYS;
+        let output = make_exec_output(exit_code, "", "", "");
+        assert!(is_likely_sandbox_denied(SandboxType::LinuxSeccomp, &output));
+    }
+}

codex-rs/core/src/exec_command/exec_command_params.rs

@@ -1,57 +0,0 @@
-use serde::Deserialize;
-use serde::Serialize;
-
-use crate::exec_command::session_id::SessionId;
-
-#[derive(Debug, Clone, Deserialize)]
-pub struct ExecCommandParams {
-    pub(crate) cmd: String,
-
-    #[serde(default = "default_yield_time")]
-    pub(crate) yield_time_ms: u64,
-
-    #[serde(default = "max_output_tokens")]
-    pub(crate) max_output_tokens: u64,
-
-    #[serde(default = "default_shell")]
-    pub(crate) shell: String,
-
-    #[serde(default = "default_login")]
-    pub(crate) login: bool,
-}
-
-fn default_yield_time() -> u64 {
-    10_000
-}
-
-fn max_output_tokens() -> u64 {
-    10_000
-}
-
-fn default_login() -> bool {
-    true
-}
-
-fn default_shell() -> String {
-    "/bin/bash".to_string()
-}
-
-#[derive(Debug, Deserialize, Serialize)]
-pub struct WriteStdinParams {
-    pub(crate) session_id: SessionId,
-    pub(crate) chars: String,
-
-    #[serde(default = "write_stdin_default_yield_time_ms")]
-    pub(crate) yield_time_ms: u64,
-
-    #[serde(default = "write_stdin_default_max_output_tokens")]
-    pub(crate) max_output_tokens: u64,
-}
-
-fn write_stdin_default_yield_time_ms() -> u64 {
-    250
-}
-
-fn write_stdin_default_max_output_tokens() -> u64 {
-    10_000
-}

codex-rs/core/src/exec_command/exec_command_session.rs

@@ -1,96 +0,0 @@
-use std::sync::Mutex as StdMutex;
-
-use tokio::sync::broadcast;
-use tokio::sync::mpsc;
-use tokio::task::JoinHandle;
-
-#[derive(Debug)]
-pub(crate) struct ExecCommandSession {
-    /// Queue for writing bytes to the process stdin (PTY master write side).
-    writer_tx: mpsc::Sender<Vec<u8>>,
-    /// Broadcast stream of output chunks read from the PTY. New subscribers
-    /// receive only chunks emitted after they subscribe.
-    output_tx: broadcast::Sender<Vec<u8>>,
-
-    /// Child killer handle for termination on drop (can signal independently
-    /// of a thread blocked in `.wait()`).
-    killer: StdMutex<Option<Box<dyn portable_pty::ChildKiller + Send + Sync>>>,
-
-    /// JoinHandle for the blocking PTY reader task.
-    reader_handle: StdMutex<Option<JoinHandle<()>>>,
-
-    /// JoinHandle for the stdin writer task.
-    writer_handle: StdMutex<Option<JoinHandle<()>>>,
-
-    /// JoinHandle for the child wait task.
-    wait_handle: StdMutex<Option<JoinHandle<()>>>,
-
-    /// Tracks whether the underlying process has exited.
-    exit_status: std::sync::Arc<std::sync::atomic::AtomicBool>,
-}
-
-impl ExecCommandSession {
-    pub(crate) fn new(
-        writer_tx: mpsc::Sender<Vec<u8>>,
-        output_tx: broadcast::Sender<Vec<u8>>,
-        killer: Box<dyn portable_pty::ChildKiller + Send + Sync>,
-        reader_handle: JoinHandle<()>,
-        writer_handle: JoinHandle<()>,
-        wait_handle: JoinHandle<()>,
-        exit_status: std::sync::Arc<std::sync::atomic::AtomicBool>,
-    ) -> (Self, broadcast::Receiver<Vec<u8>>) {
-        let initial_output_rx = output_tx.subscribe();
-        (
-            Self {
-                writer_tx,
-                output_tx,
-                killer: StdMutex::new(Some(killer)),
-                reader_handle: StdMutex::new(Some(reader_handle)),
-                writer_handle: StdMutex::new(Some(writer_handle)),
-                wait_handle: StdMutex::new(Some(wait_handle)),
-                exit_status,
-            },
-            initial_output_rx,
-        )
-    }
-
-    pub(crate) fn writer_sender(&self) -> mpsc::Sender<Vec<u8>> {
-        self.writer_tx.clone()
-    }
-
-    pub(crate) fn output_receiver(&self) -> broadcast::Receiver<Vec<u8>> {
-        self.output_tx.subscribe()
-    }
-
-    pub(crate) fn has_exited(&self) -> bool {
-        self.exit_status.load(std::sync::atomic::Ordering::SeqCst)
-    }
-}
-
-impl Drop for ExecCommandSession {
-    fn drop(&mut self) {
-        // Best-effort: terminate child first so blocking tasks can complete.
-        if let Ok(mut killer_opt) = self.killer.lock()
-            && let Some(mut killer) = killer_opt.take()
-        {
-            let _ = killer.kill();
-        }
-
-        // Abort background tasks; they may already have exited after kill.
-        if let Ok(mut h) = self.reader_handle.lock()
-            && let Some(handle) = h.take()
-        {
-            handle.abort();
-        }
-        if let Ok(mut h) = self.writer_handle.lock()
-            && let Some(handle) = h.take()
-        {
-            handle.abort();
-        }
-        if let Ok(mut h) = self.wait_handle.lock()
-            && let Some(handle) = h.take()
-        {
-            handle.abort();
-        }
-    }
-}

codex-rs/core/src/exec_command/mod.rs

@@ -1,14 +0,0 @@
-mod exec_command_params;
-mod exec_command_session;
-mod responses_api;
-mod session_id;
-mod session_manager;
-
-pub use exec_command_params::ExecCommandParams;
-pub use exec_command_params::WriteStdinParams;
-pub(crate) use exec_command_session::ExecCommandSession;
-pub use responses_api::EXEC_COMMAND_TOOL_NAME;
-pub use responses_api::WRITE_STDIN_TOOL_NAME;
-pub use responses_api::create_exec_command_tool_for_responses_api;
-pub use responses_api::create_write_stdin_tool_for_responses_api;
-pub use session_manager::SessionManager as ExecSessionManager;

codex-rs/core/src/exec_command/responses_api.rs

@@ -1,98 +0,0 @@
-use std::collections::BTreeMap;
-
-use crate::openai_tools::JsonSchema;
-use crate::openai_tools::ResponsesApiTool;
-
-pub const EXEC_COMMAND_TOOL_NAME: &str = "exec_command";
-pub const WRITE_STDIN_TOOL_NAME: &str = "write_stdin";
-
-pub fn create_exec_command_tool_for_responses_api() -> ResponsesApiTool {
-    let mut properties = BTreeMap::<String, JsonSchema>::new();
-    properties.insert(
-        "cmd".to_string(),
-        JsonSchema::String {
-            description: Some("The shell command to execute.".to_string()),
-        },
-    );
-    properties.insert(
-        "yield_time_ms".to_string(),
-        JsonSchema::Number {
-            description: Some("The maximum time in milliseconds to wait for output.".to_string()),
-        },
-    );
-    properties.insert(
-        "max_output_tokens".to_string(),
-        JsonSchema::Number {
-            description: Some("The maximum number of tokens to output.".to_string()),
-        },
-    );
-    properties.insert(
-        "shell".to_string(),
-        JsonSchema::String {
-            description: Some("The shell to use. Defaults to \"/bin/bash\".".to_string()),
-        },
-    );
-    properties.insert(
-        "login".to_string(),
-        JsonSchema::Boolean {
-            description: Some(
-                "Whether to run the command as a login shell. Defaults to true.".to_string(),
-            ),
-        },
-    );
-
-    ResponsesApiTool {
-        name: EXEC_COMMAND_TOOL_NAME.to_owned(),
-        description: r#"Execute shell commands on the local machine with streaming output."#
-            .to_string(),
-        strict: false,
-        parameters: JsonSchema::Object {
-            properties,
-            required: Some(vec!["cmd".to_string()]),
-            additional_properties: Some(false),
-        },
-    }
-}
-
-pub fn create_write_stdin_tool_for_responses_api() -> ResponsesApiTool {
-    let mut properties = BTreeMap::<String, JsonSchema>::new();
-    properties.insert(
-        "session_id".to_string(),
-        JsonSchema::Number {
-            description: Some("The ID of the exec_command session.".to_string()),
-        },
-    );
-    properties.insert(
-        "chars".to_string(),
-        JsonSchema::String {
-            description: Some("The characters to write to stdin.".to_string()),
-        },
-    );
-    properties.insert(
-        "yield_time_ms".to_string(),
-        JsonSchema::Number {
-            description: Some(
-                "The maximum time in milliseconds to wait for output after writing.".to_string(),
-            ),
-        },
-    );
-    properties.insert(
-        "max_output_tokens".to_string(),
-        JsonSchema::Number {
-            description: Some("The maximum number of tokens to output.".to_string()),
-        },
-    );
-
-    ResponsesApiTool {
-        name: WRITE_STDIN_TOOL_NAME.to_owned(),
-        description: r#"Write characters to an exec session's stdin. Returns all stdout+stderr received within yield_time_ms.
-Can write control characters (\u0003 for Ctrl-C), or an empty string to just poll stdout+stderr."#
-            .to_string(),
-        strict: false,
-        parameters: JsonSchema::Object {
-            properties,
-            required: Some(vec!["session_id".to_string(), "chars".to_string()]),
-            additional_properties: Some(false),
-        },
-    }
-}

codex-rs/core/src/exec_command/session_id.rs

@@ -1,5 +0,0 @@
-use serde::Deserialize;
-use serde::Serialize;
-
-#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash, Serialize, Deserialize)]
-pub(crate) struct SessionId(pub u32);

codex-rs/core/src/exec_command/session_manager.rs

@@ -1,506 +0,0 @@
-use std::collections::HashMap;
-use std::io::ErrorKind;
-use std::io::Read;
-use std::sync::Arc;
-use std::sync::Mutex as StdMutex;
-use std::sync::atomic::AtomicBool;
-use std::sync::atomic::AtomicU32;
-
-use portable_pty::CommandBuilder;
-use portable_pty::PtySize;
-use portable_pty::native_pty_system;
-use tokio::sync::Mutex;
-use tokio::sync::mpsc;
-use tokio::sync::oneshot;
-use tokio::time::Duration;
-use tokio::time::Instant;
-use tokio::time::timeout;
-
-use crate::exec_command::exec_command_params::ExecCommandParams;
-use crate::exec_command::exec_command_params::WriteStdinParams;
-use crate::exec_command::exec_command_session::ExecCommandSession;
-use crate::exec_command::session_id::SessionId;
-use crate::truncate::truncate_middle;
-
-#[derive(Debug, Default)]
-pub struct SessionManager {
-    next_session_id: AtomicU32,
-    sessions: Mutex<HashMap<SessionId, ExecCommandSession>>,
-}
-
-#[derive(Debug)]
-pub struct ExecCommandOutput {
-    wall_time: Duration,
-    exit_status: ExitStatus,
-    original_token_count: Option<u64>,
-    output: String,
-}
-
-impl ExecCommandOutput {
-    pub(crate) fn to_text_output(&self) -> String {
-        let wall_time_secs = self.wall_time.as_secs_f32();
-        let termination_status = match self.exit_status {
-            ExitStatus::Exited(code) => format!("Process exited with code {code}"),
-            ExitStatus::Ongoing(session_id) => {
-                format!("Process running with session ID {}", session_id.0)
-            }
-        };
-        let truncation_status = match self.original_token_count {
-            Some(tokens) => {
-                format!("\nWarning: truncated output (original token count: {tokens})")
-            }
-            None => "".to_string(),
-        };
-        format!(
-            r#"Wall time: {wall_time_secs:.3} seconds
-{termination_status}{truncation_status}
-Output:
-{output}"#,
-            output = self.output
-        )
-    }
-}
-
-#[derive(Debug)]
-pub enum ExitStatus {
-    Exited(i32),
-    Ongoing(SessionId),
-}
-
-impl SessionManager {
-    /// Processes the request and is required to send a response via `outgoing`.
-    pub async fn handle_exec_command_request(
-        &self,
-        params: ExecCommandParams,
-    ) -> Result<ExecCommandOutput, String> {
-        // Allocate a session id.
-        let session_id = SessionId(
-            self.next_session_id
-                .fetch_add(1, std::sync::atomic::Ordering::SeqCst),
-        );
-
-        let (session, mut output_rx, mut exit_rx) = create_exec_command_session(params.clone())
-            .await
-            .map_err(|err| {
-                format!(
-                    "failed to create exec command session for session id {}: {err}",
-                    session_id.0
-                )
-            })?;
-
-        // Insert into session map.
-        self.sessions.lock().await.insert(session_id, session);
-
-        // Collect output until either timeout expires or process exits.
-        // Do not cap during collection; truncate at the end if needed.
-        // Use a modest initial capacity to avoid large preallocation.
-        let cap_bytes_u64 = params.max_output_tokens.saturating_mul(4);
-        let cap_bytes: usize = cap_bytes_u64.min(usize::MAX as u64) as usize;
-        let mut collected: Vec<u8> = Vec::with_capacity(4096);
-
-        let start_time = Instant::now();
-        let deadline = start_time + Duration::from_millis(params.yield_time_ms);
-        let mut exit_code: Option<i32> = None;
-
-        loop {
-            if Instant::now() >= deadline {
-                break;
-            }
-            let remaining = deadline.saturating_duration_since(Instant::now());
-            tokio::select! {
-                biased;
-                exit = &mut exit_rx => {
-                    exit_code = exit.ok();
-                    // Small grace period to pull remaining buffered output
-                    let grace_deadline = Instant::now() + Duration::from_millis(25);
-                    while Instant::now() < grace_deadline {
-                        match timeout(Duration::from_millis(1), output_rx.recv()).await {
-                            Ok(Ok(chunk)) => {
-                                collected.extend_from_slice(&chunk);
-                            }
-                            Ok(Err(tokio::sync::broadcast::error::RecvError::Lagged(_))) => {
-                                // Skip missed messages; keep trying within grace period.
-                                continue;
-                            }
-                            Ok(Err(tokio::sync::broadcast::error::RecvError::Closed)) => break,
-                            Err(_) => break,
-                        }
-                    }
-                    break;
-                }
-                chunk = timeout(remaining, output_rx.recv()) => {
-                    match chunk {
-                        Ok(Ok(chunk)) => {
-                            collected.extend_from_slice(&chunk);
-                        }
-                        Ok(Err(tokio::sync::broadcast::error::RecvError::Lagged(_))) => {
-                            // Skip missed messages; continue collecting fresh output.
-                        }
-                        Ok(Err(tokio::sync::broadcast::error::RecvError::Closed)) => { break; }
-                        Err(_) => { break; }
-                    }
-                }
-            }
-        }
-
-        let output = String::from_utf8_lossy(&collected).to_string();
-
-        let exit_status = if let Some(code) = exit_code {
-            ExitStatus::Exited(code)
-        } else {
-            ExitStatus::Ongoing(session_id)
-        };
-
-        // If output exceeds cap, truncate the middle and record original token estimate.
-        let (output, original_token_count) = truncate_middle(&output, cap_bytes);
-        Ok(ExecCommandOutput {
-            wall_time: Instant::now().duration_since(start_time),
-            exit_status,
-            original_token_count,
-            output,
-        })
-    }
-
-    /// Write characters to a session's stdin and collect combined output for up to `yield_time_ms`.
-    pub async fn handle_write_stdin_request(
-        &self,
-        params: WriteStdinParams,
-    ) -> Result<ExecCommandOutput, String> {
-        let WriteStdinParams {
-            session_id,
-            chars,
-            yield_time_ms,
-            max_output_tokens,
-        } = params;
-
-        // Grab handles without holding the sessions lock across await points.
-        let (writer_tx, mut output_rx) = {
-            let sessions = self.sessions.lock().await;
-            match sessions.get(&session_id) {
-                Some(session) => (session.writer_sender(), session.output_receiver()),
-                None => {
-                    return Err(format!("unknown session id {}", session_id.0));
-                }
-            }
-        };
-
-        // Write stdin if provided.
-        if !chars.is_empty() && writer_tx.send(chars.into_bytes()).await.is_err() {
-            return Err("failed to write to stdin".to_string());
-        }
-
-        // Collect output up to yield_time_ms, truncating to max_output_tokens bytes.
-        let mut collected: Vec<u8> = Vec::with_capacity(4096);
-        let start_time = Instant::now();
-        let deadline = start_time + Duration::from_millis(yield_time_ms);
-        loop {
-            let now = Instant::now();
-            if now >= deadline {
-                break;
-            }
-            let remaining = deadline - now;
-            match timeout(remaining, output_rx.recv()).await {
-                Ok(Ok(chunk)) => {
-                    // Collect all output within the time budget; truncate at the end.
-                    collected.extend_from_slice(&chunk);
-                }
-                Ok(Err(tokio::sync::broadcast::error::RecvError::Lagged(_))) => {
-                    // Skip missed messages; continue collecting fresh output.
-                }
-                Ok(Err(tokio::sync::broadcast::error::RecvError::Closed)) => break,
-                Err(_) => break, // timeout
-            }
-        }
-
-        // Return structured output, truncating middle if over cap.
-        let output = String::from_utf8_lossy(&collected).to_string();
-        let cap_bytes_u64 = max_output_tokens.saturating_mul(4);
-        let cap_bytes: usize = cap_bytes_u64.min(usize::MAX as u64) as usize;
-        let (output, original_token_count) = truncate_middle(&output, cap_bytes);
-        Ok(ExecCommandOutput {
-            wall_time: Instant::now().duration_since(start_time),
-            exit_status: ExitStatus::Ongoing(session_id),
-            original_token_count,
-            output,
-        })
-    }
-}
-
-/// Spawn PTY and child process per spawn_exec_command_session logic.
-async fn create_exec_command_session(
-    params: ExecCommandParams,
-) -> anyhow::Result<(
-    ExecCommandSession,
-    tokio::sync::broadcast::Receiver<Vec<u8>>,
-    oneshot::Receiver<i32>,
-)> {
-    let ExecCommandParams {
-        cmd,
-        yield_time_ms: _,
-        max_output_tokens: _,
-        shell,
-        login,
-    } = params;
-
-    // Use the native pty implementation for the system
-    let pty_system = native_pty_system();
-
-    // Create a new pty
-    let pair = pty_system.openpty(PtySize {
-        rows: 24,
-        cols: 80,
-        pixel_width: 0,
-        pixel_height: 0,
-    })?;
-
-    // Spawn a shell into the pty
-    let mut command_builder = CommandBuilder::new(shell);
-    let shell_mode_opt = if login { "-lc" } else { "-c" };
-    command_builder.arg(shell_mode_opt);
-    command_builder.arg(cmd);
-
-    let mut child = pair.slave.spawn_command(command_builder)?;
-    // Obtain a killer that can signal the process independently of `.wait()`.
-    let killer = child.clone_killer();
-
-    // Channel to forward write requests to the PTY writer.
-    let (writer_tx, mut writer_rx) = mpsc::channel::<Vec<u8>>(128);
-    // Broadcast for streaming PTY output to readers: subscribers receive from subscription time.
-    let (output_tx, _) = tokio::sync::broadcast::channel::<Vec<u8>>(256);
-    // Reader task: drain PTY and forward chunks to output channel.
-    let mut reader = pair.master.try_clone_reader()?;
-    let output_tx_clone = output_tx.clone();
-    let reader_handle = tokio::task::spawn_blocking(move || {
-        let mut buf = [0u8; 8192];
-        loop {
-            match reader.read(&mut buf) {
-                Ok(0) => break, // EOF
-                Ok(n) => {
-                    // Forward to broadcast; best-effort if there are subscribers.
-                    let _ = output_tx_clone.send(buf[..n].to_vec());
-                }
-                Err(ref e) if e.kind() == ErrorKind::Interrupted => {
-                    // Retry on EINTR
-                    continue;
-                }
-                Err(ref e) if e.kind() == ErrorKind::WouldBlock => {
-                    // We're in a blocking thread; back off briefly and retry.
-                    std::thread::sleep(Duration::from_millis(5));
-                    continue;
-                }
-                Err(_) => break,
-            }
-        }
-    });
-
-    // Writer task: apply stdin writes to the PTY writer.
-    let writer = pair.master.take_writer()?;
-    let writer = Arc::new(StdMutex::new(writer));
-    let writer_handle = tokio::spawn({
-        let writer = writer.clone();
-        async move {
-            while let Some(bytes) = writer_rx.recv().await {
-                let writer = writer.clone();
-                // Perform blocking write on a blocking thread.
-                let _ = tokio::task::spawn_blocking(move || {
-                    if let Ok(mut guard) = writer.lock() {
-                        use std::io::Write;
-                        let _ = guard.write_all(&bytes);
-                        let _ = guard.flush();
-                    }
-                })
-                .await;
-            }
-        }
-    });
-
-    // Keep the child alive until it exits, then signal exit code.
-    let (exit_tx, exit_rx) = oneshot::channel::<i32>();
-    let exit_status = Arc::new(AtomicBool::new(false));
-    let wait_exit_status = exit_status.clone();
-    let wait_handle = tokio::task::spawn_blocking(move || {
-        let code = match child.wait() {
-            Ok(status) => status.exit_code() as i32,
-            Err(_) => -1,
-        };
-        wait_exit_status.store(true, std::sync::atomic::Ordering::SeqCst);
-        let _ = exit_tx.send(code);
-    });
-
-    // Create and store the session with channels.
-    let (session, initial_output_rx) = ExecCommandSession::new(
-        writer_tx,
-        output_tx,
-        killer,
-        reader_handle,
-        writer_handle,
-        wait_handle,
-        exit_status,
-    );
-    Ok((session, initial_output_rx, exit_rx))
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use crate::exec_command::session_id::SessionId;
-
-    /// Test that verifies that [`SessionManager::handle_exec_command_request()`]
-    /// and [`SessionManager::handle_write_stdin_request()`] work as expected
-    /// in the presence of a process that never terminates (but produces
-    /// output continuously).
-    #[cfg(unix)]
-    #[allow(clippy::print_stderr)]
-    #[tokio::test(flavor = "multi_thread", worker_threads = 4)]
-    async fn session_manager_streams_and_truncates_from_now() {
-        use crate::exec_command::exec_command_params::ExecCommandParams;
-        use crate::exec_command::exec_command_params::WriteStdinParams;
-        use tokio::time::sleep;
-
-        let session_manager = SessionManager::default();
-        // Long-running loop that prints an increasing counter every ~100ms.
-        // Use Python for a portable, reliable sleep across shells/PTYs.
-        let cmd = r#"python3 - <<'PY'
-import sys, time
-count = 0
-while True:
-    print(count)
-    sys.stdout.flush()
-    count += 100
-    time.sleep(0.1)
-PY"#
-        .to_string();
-
-        // Start the session and collect ~3s of output.
-        let params = ExecCommandParams {
-            cmd,
-            yield_time_ms: 3_000,
-            max_output_tokens: 1_000, // large enough to avoid truncation here
-            shell: "/bin/bash".to_string(),
-            login: false,
-        };
-        let initial_output = match session_manager
-            .handle_exec_command_request(params.clone())
-            .await
-        {
-            Ok(v) => v,
-            Err(e) => {
-                // PTY may be restricted in some sandboxes; skip in that case.
-                if e.contains("openpty") || e.contains("Operation not permitted") {
-                    eprintln!("skipping test due to restricted PTY: {e}");
-                    return;
-                }
-                panic!("exec request failed unexpectedly: {e}");
-            }
-        };
-        eprintln!("initial output: {initial_output:?}");
-
-        // Should be ongoing (we launched a never-ending loop).
-        let session_id = match initial_output.exit_status {
-            ExitStatus::Ongoing(id) => id,
-            _ => panic!("expected ongoing session"),
-        };
-
-        // Parse the numeric lines and get the max observed value in the first window.
-        let first_nums = extract_monotonic_numbers(&initial_output.output);
-        assert!(
-            !first_nums.is_empty(),
-            "expected some output from first window"
-        );
-        let first_max = *first_nums.iter().max().unwrap();
-
-        // Wait ~4s so counters progress while we're not reading.
-        sleep(Duration::from_millis(4_000)).await;
-
-        // Now read ~3s of output "from now" only.
-        // Use a small token cap so truncation occurs and we test middle truncation.
-        let write_params = WriteStdinParams {
-            session_id,
-            chars: String::new(),
-            yield_time_ms: 3_000,
-            max_output_tokens: 16, // 16 tokens ~= 64 bytes -> likely truncation
-        };
-        let second = session_manager
-            .handle_write_stdin_request(write_params)
-            .await
-            .expect("write stdin should succeed");
-
-        // Verify truncation metadata and size bound (cap is tokens*4 bytes).
-        assert!(second.original_token_count.is_some());
-        let cap_bytes = (16u64 * 4) as usize;
-        assert!(second.output.len() <= cap_bytes);
-        // New middle marker should be present.
-        assert!(
-            second.output.contains("tokens truncated") && second.output.contains('…'),
-            "expected truncation marker in output, got: {}",
-            second.output
-        );
-
-        // Minimal freshness check: the earliest number we see in the second window
-        // should be significantly larger than the last from the first window.
-        let second_nums = extract_monotonic_numbers(&second.output);
-        assert!(
-            !second_nums.is_empty(),
-            "expected some numeric output from second window"
-        );
-        let second_min = *second_nums.iter().min().unwrap();
-
-        // We slept 4 seconds (~40 ticks at 100ms/tick, each +100), so expect
-        // an increase of roughly 4000 or more. Allow a generous margin.
-        assert!(
-            second_min >= first_max + 2000,
-            "second_min={second_min} first_max={first_max}",
-        );
-    }
-
-    #[cfg(unix)]
-    fn extract_monotonic_numbers(s: &str) -> Vec<i64> {
-        s.lines()
-            .filter_map(|line| {
-                if !line.is_empty()
-                    && line.chars().all(|c| c.is_ascii_digit())
-                    && let Ok(n) = line.parse::<i64>()
-                {
-                    // Our generator increments by 100; ignore spurious fragments.
-                    if n % 100 == 0 {
-                        return Some(n);
-                    }
-                }
-                None
-            })
-            .collect()
-    }
-
-    #[test]
-    fn to_text_output_exited_no_truncation() {
-        let out = ExecCommandOutput {
-            wall_time: Duration::from_millis(1234),
-            exit_status: ExitStatus::Exited(0),
-            original_token_count: None,
-            output: "hello".to_string(),
-        };
-        let text = out.to_text_output();
-        let expected = r#"Wall time: 1.234 seconds
-Process exited with code 0
-Output:
-hello"#;
-        assert_eq!(expected, text);
-    }
-
-    #[test]
-    fn to_text_output_ongoing_with_truncation() {
-        let out = ExecCommandOutput {
-            wall_time: Duration::from_millis(500),
-            exit_status: ExitStatus::Ongoing(SessionId(42)),
-            original_token_count: Some(1000),
-            output: "abc".to_string(),
-        };
-        let text = out.to_text_output();
-        let expected = r#"Wall time: 0.500 seconds
-Process running with session ID 42
-Warning: truncated output (original token count: 1000)
-Output:
-abc"#;
-        assert_eq!(expected, text);
-    }
-}

codex-rs/core/src/executor/backends.rs

@@ -1,101 +0,0 @@
-use std::collections::HashMap;
-use std::env;
-
-use async_trait::async_trait;
-
-use crate::CODEX_APPLY_PATCH_ARG1;
-use crate::apply_patch::ApplyPatchExec;
-use crate::exec::ExecParams;
-use crate::function_tool::FunctionCallError;
-
-pub(crate) enum ExecutionMode {
-    Shell,
-    ApplyPatch(ApplyPatchExec),
-}
-
-#[async_trait]
-/// Backend-specific hooks that prepare and post-process execution requests for a
-/// given [`ExecutionMode`].
-pub(crate) trait ExecutionBackend: Send + Sync {
-    fn prepare(
-        &self,
-        params: ExecParams,
-        // Required for downcasting the apply_patch.
-        mode: &ExecutionMode,
-    ) -> Result<ExecParams, FunctionCallError>;
-
-    fn stream_stdout(&self, _mode: &ExecutionMode) -> bool {
-        true
-    }
-}
-
-static SHELL_BACKEND: ShellBackend = ShellBackend;
-static APPLY_PATCH_BACKEND: ApplyPatchBackend = ApplyPatchBackend;
-
-pub(crate) fn backend_for_mode(mode: &ExecutionMode) -> &'static dyn ExecutionBackend {
-    match mode {
-        ExecutionMode::Shell => &SHELL_BACKEND,
-        ExecutionMode::ApplyPatch(_) => &APPLY_PATCH_BACKEND,
-    }
-}
-
-struct ShellBackend;
-
-#[async_trait]
-impl ExecutionBackend for ShellBackend {
-    fn prepare(
-        &self,
-        params: ExecParams,
-        mode: &ExecutionMode,
-    ) -> Result<ExecParams, FunctionCallError> {
-        match mode {
-            ExecutionMode::Shell => Ok(params),
-            _ => Err(FunctionCallError::RespondToModel(
-                "shell backend invoked with non-shell mode".to_string(),
-            )),
-        }
-    }
-}
-
-struct ApplyPatchBackend;
-
-#[async_trait]
-impl ExecutionBackend for ApplyPatchBackend {
-    fn prepare(
-        &self,
-        params: ExecParams,
-        mode: &ExecutionMode,
-    ) -> Result<ExecParams, FunctionCallError> {
-        match mode {
-            ExecutionMode::ApplyPatch(exec) => {
-                let path_to_codex = env::current_exe()
-                    .ok()
-                    .map(|p| p.to_string_lossy().to_string())
-                    .ok_or_else(|| {
-                        FunctionCallError::RespondToModel(
-                            "failed to determine path to codex executable".to_string(),
-                        )
-                    })?;
-
-                let patch = exec.action.patch.clone();
-                Ok(ExecParams {
-                    command: vec![path_to_codex, CODEX_APPLY_PATCH_ARG1.to_string(), patch],
-                    cwd: exec.action.cwd.clone(),
-                    timeout_ms: params.timeout_ms,
-                    // Run apply_patch with a minimal environment for determinism and to
-                    // avoid leaking host environment variables into the patch process.
-                    env: HashMap::new(),
-                    with_escalated_permissions: params.with_escalated_permissions,
-                    justification: params.justification,
-                })
-            }
-            ExecutionMode::Shell => Err(FunctionCallError::RespondToModel(
-                "apply_patch backend invoked without patch context".to_string(),
-            )),
-        }
-    }
-
-    fn stream_stdout(&self, _mode: &ExecutionMode) -> bool {
-        false
-    }
-}

codex-rs/core/src/executor/cache.rs

@@ -1,51 +0,0 @@
-use std::collections::HashSet;
-use std::sync::Arc;
-use std::sync::Mutex;
-
-#[derive(Clone, Debug, Default)]
-/// Thread-safe store of user approvals so repeated commands can reuse
-/// previously granted trust.
-pub(crate) struct ApprovalCache {
-    inner: Arc<Mutex<HashSet<Vec<String>>>>,
-}
-
-impl ApprovalCache {
-    pub(crate) fn insert(&self, command: Vec<String>) {
-        if command.is_empty() {
-            return;
-        }
-        if let Ok(mut guard) = self.inner.lock() {
-            guard.insert(command);
-        }
-    }
-
-    pub(crate) fn snapshot(&self) -> HashSet<Vec<String>> {
-        self.inner.lock().map(|g| g.clone()).unwrap_or_default()
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use pretty_assertions::assert_eq;
-
-    #[test]
-    fn insert_ignores_empty_and_dedupes() {
-        let cache = ApprovalCache::default();
-
-        // Empty should be ignored
-        cache.insert(vec![]);
-        assert!(cache.snapshot().is_empty());
-
-        // Insert a command and verify snapshot contains it
-        let cmd = vec!["foo".to_string(), "bar".to_string()];
-        cache.insert(cmd.clone());
-        let snap1 = cache.snapshot();
-        assert!(snap1.contains(&cmd));
-
-        // Reinserting should not create duplicates
-        cache.insert(cmd);
-        let snap2 = cache.snapshot();
-        assert_eq!(snap1, snap2);
-    }
-}

codex-rs/core/src/executor/mod.rs

@@ -1,64 +0,0 @@
-mod backends;
-mod cache;
-mod runner;
-mod sandbox;
-
-pub(crate) use backends::ExecutionMode;
-pub(crate) use runner::ExecutionRequest;
-pub(crate) use runner::Executor;
-pub(crate) use runner::ExecutorConfig;
-pub(crate) use runner::normalize_exec_result;
-
-pub(crate) mod linkers {
-    use crate::codex::ExecCommandContext;
-    use crate::exec::ExecParams;
-    use crate::exec::StdoutStream;
-    use crate::executor::backends::ExecutionMode;
-    use crate::executor::runner::ExecutionRequest;
-
-    pub struct PreparedExec {
-        pub(crate) context: ExecCommandContext,
-        pub(crate) request: ExecutionRequest,
-    }
-
-    impl PreparedExec {
-        pub fn new(
-            context: ExecCommandContext,
-            params: ExecParams,
-            approval_command: Vec<String>,
-            mode: ExecutionMode,
-            stdout_stream: Option<StdoutStream>,
-            use_shell_profile: bool,
-        ) -> Self {
-            let request = ExecutionRequest {
-                params,
-                approval_command,
-                mode,
-                stdout_stream,
-                use_shell_profile,
-            };
-
-            Self { context, request }
-        }
-    }
-}
-
-pub mod errors {
-    use crate::error::CodexErr;
-    use crate::function_tool::FunctionCallError;
-    use thiserror::Error;
-
-    #[derive(Debug, Error)]
-    pub enum ExecError {
-        #[error(transparent)]
-        Function(#[from] FunctionCallError),
-        #[error(transparent)]
-        Codex(#[from] CodexErr),
-    }
-
-    impl ExecError {
-        pub(crate) fn rejection(msg: impl Into<String>) -> Self {
-            FunctionCallError::RespondToModel(msg.into()).into()
-        }
-    }
-}

codex-rs/core/src/executor/runner.rs

@@ -1,387 +0,0 @@
-use std::path::PathBuf;
-use std::sync::Arc;
-use std::sync::RwLock;
-use std::time::Duration;
-
-use super::backends::ExecutionMode;
-use super::backends::backend_for_mode;
-use super::cache::ApprovalCache;
-use crate::codex::ExecCommandContext;
-use crate::codex::Session;
-use crate::error::CodexErr;
-use crate::error::SandboxErr;
-use crate::error::get_error_message_ui;
-use crate::exec::ExecParams;
-use crate::exec::ExecToolCallOutput;
-use crate::exec::SandboxType;
-use crate::exec::StdoutStream;
-use crate::exec::StreamOutput;
-use crate::exec::process_exec_tool_call;
-use crate::executor::errors::ExecError;
-use crate::executor::sandbox::select_sandbox;
-use crate::function_tool::FunctionCallError;
-use crate::protocol::AskForApproval;
-use crate::protocol::ReviewDecision;
-use crate::protocol::SandboxPolicy;
-use crate::shell;
-use codex_otel::otel_event_manager::ToolDecisionSource;
-
-#[derive(Clone, Debug)]
-pub(crate) struct ExecutorConfig {
-    pub(crate) sandbox_policy: SandboxPolicy,
-    pub(crate) sandbox_cwd: PathBuf,
-    codex_linux_sandbox_exe: Option<PathBuf>,
-}
-
-impl ExecutorConfig {
-    pub(crate) fn new(
-        sandbox_policy: SandboxPolicy,
-        sandbox_cwd: PathBuf,
-        codex_linux_sandbox_exe: Option<PathBuf>,
-    ) -> Self {
-        Self {
-            sandbox_policy,
-            sandbox_cwd,
-            codex_linux_sandbox_exe,
-        }
-    }
-}
-
-/// Coordinates sandbox selection, backend-specific preparation, and command
-/// execution for tool calls requested by the model.
-pub(crate) struct Executor {
-    approval_cache: ApprovalCache,
-    config: Arc<RwLock<ExecutorConfig>>,
-}
-
-impl Executor {
-    pub(crate) fn new(config: ExecutorConfig) -> Self {
-        Self {
-            approval_cache: ApprovalCache::default(),
-            config: Arc::new(RwLock::new(config)),
-        }
-    }
-
-    /// Updates the sandbox policy and working directory used for future
-    /// executions without recreating the executor.
-    pub(crate) fn update_environment(&self, sandbox_policy: SandboxPolicy, sandbox_cwd: PathBuf) {
-        if let Ok(mut cfg) = self.config.write() {
-            cfg.sandbox_policy = sandbox_policy;
-            cfg.sandbox_cwd = sandbox_cwd;
-        }
-    }
-
-    /// Runs a prepared execution request end-to-end: prepares parameters, decides on
-    /// sandbox placement (prompting the user when necessary), launches the command,
-    /// and lets the backend post-process the final output.
-    pub(crate) async fn run(
-        &self,
-        mut request: ExecutionRequest,
-        session: &Session,
-        approval_policy: AskForApproval,
-        context: &ExecCommandContext,
-    ) -> Result<ExecToolCallOutput, ExecError> {
-        if matches!(request.mode, ExecutionMode::Shell) {
-            request.params =
-                maybe_translate_shell_command(request.params, session, request.use_shell_profile);
-        }
-
-        // Step 1: Normalise parameters via the selected backend.
-        let backend = backend_for_mode(&request.mode);
-        let stdout_stream = if backend.stream_stdout(&request.mode) {
-            request.stdout_stream.clone()
-        } else {
-            None
-        };
-        request.params = backend
-            .prepare(request.params, &request.mode)
-            .map_err(ExecError::from)?;
-
-        // Step 2: Snapshot sandbox configuration so it stays stable for this run.
-        let config = self
-            .config
-            .read()
-            .map_err(|_| ExecError::rejection("executor config poisoned"))?
-            .clone();
-
-        // Step 3: Decide sandbox placement, prompting for approval when needed.
-        let sandbox_decision = select_sandbox(
-            &request,
-            approval_policy,
-            self.approval_cache.snapshot(),
-            &config,
-            session,
-            &context.sub_id,
-            &context.call_id,
-            &context.otel_event_manager,
-        )
-        .await?;
-        if sandbox_decision.record_session_approval {
-            self.approval_cache.insert(request.approval_command.clone());
-        }
-
-        // Step 4: Launch the command within the chosen sandbox.
-        let first_attempt = self
-            .spawn(
-                request.params.clone(),
-                sandbox_decision.initial_sandbox,
-                &config,
-                stdout_stream.clone(),
-            )
-            .await;
-
-        // Step 5: Handle sandbox outcomes, optionally escalating to an unsandboxed retry.
-        match first_attempt {
-            Ok(output) => Ok(output),
-            Err(CodexErr::Sandbox(SandboxErr::Timeout { output })) => {
-                Err(CodexErr::Sandbox(SandboxErr::Timeout { output }).into())
-            }
-            Err(CodexErr::Sandbox(error)) => {
-                if sandbox_decision.escalate_on_failure {
-                    self.retry_without_sandbox(
-                        &request,
-                        &config,
-                        session,
-                        context,
-                        stdout_stream,
-                        error,
-                    )
-                    .await
-                } else {
-                    Err(ExecError::rejection(format!(
-                        "failed in sandbox {:?} with execution error: {error:?}",
-                        sandbox_decision.initial_sandbox
-                    )))
-                }
-            }
-            Err(err) => Err(err.into()),
-        }
-    }
-
-    /// Fallback path invoked when a sandboxed run is denied so the user can
-    /// approve rerunning without isolation.
-    async fn retry_without_sandbox(
-        &self,
-        request: &ExecutionRequest,
-        config: &ExecutorConfig,
-        session: &Session,
-        context: &ExecCommandContext,
-        stdout_stream: Option<StdoutStream>,
-        sandbox_error: SandboxErr,
-    ) -> Result<ExecToolCallOutput, ExecError> {
-        session
-            .notify_background_event(
-                &context.sub_id,
-                format!("Execution failed: {sandbox_error}"),
-            )
-            .await;
-        let decision = session
-            .request_command_approval(
-                context.sub_id.to_string(),
-                context.call_id.to_string(),
-                request.approval_command.clone(),
-                request.params.cwd.clone(),
-                Some("command failed; retry without sandbox?".to_string()),
-            )
-            .await;
-
-        context.otel_event_manager.tool_decision(
-            &context.tool_name,
-            &context.call_id,
-            decision,
-            ToolDecisionSource::User,
-        );
-        match decision {
-            ReviewDecision::Approved | ReviewDecision::ApprovedForSession => {
-                if matches!(decision, ReviewDecision::ApprovedForSession) {
-                    self.approval_cache.insert(request.approval_command.clone());
-                }
-                session
-                    .notify_background_event(&context.sub_id, "retrying command without sandbox")
-                    .await;
-
-                let retry_output = self
-                    .spawn(
-                        request.params.clone(),
-                        SandboxType::None,
-                        config,
-                        stdout_stream,
-                    )
-                    .await?;
-
-                Ok(retry_output)
-            }
-            ReviewDecision::Denied | ReviewDecision::Abort => {
-                Err(ExecError::rejection("exec command rejected by user"))
-            }
-        }
-    }
-
-    async fn spawn(
-        &self,
-        params: ExecParams,
-        sandbox: SandboxType,
-        config: &ExecutorConfig,
-        stdout_stream: Option<StdoutStream>,
-    ) -> Result<ExecToolCallOutput, CodexErr> {
-        process_exec_tool_call(
-            params,
-            sandbox,
-            &config.sandbox_policy,
-            &config.sandbox_cwd,
-            &config.codex_linux_sandbox_exe,
-            stdout_stream,
-        )
-        .await
-    }
-}
-
-fn maybe_translate_shell_command(
-    params: ExecParams,
-    session: &Session,
-    use_shell_profile: bool,
-) -> ExecParams {
-    let should_translate =
-        matches!(session.user_shell(), shell::Shell::PowerShell(_)) || use_shell_profile;
-
-    if should_translate
-        && let Some(command) = session
-            .user_shell()
-            .format_default_shell_invocation(params.command.clone())
-    {
-        return ExecParams { command, ..params };
-    }
-
-    params
-}
-
-pub(crate) struct ExecutionRequest {
-    pub params: ExecParams,
-    pub approval_command: Vec<String>,
-    pub mode: ExecutionMode,
-    pub stdout_stream: Option<StdoutStream>,
-    pub use_shell_profile: bool,
-}
-
-pub(crate) struct NormalizedExecOutput<'a> {
-    borrowed: Option<&'a ExecToolCallOutput>,
-    synthetic: Option<ExecToolCallOutput>,
-}
-
-impl<'a> NormalizedExecOutput<'a> {
-    pub(crate) fn event_output(&'a self) -> &'a ExecToolCallOutput {
-        match (self.borrowed, self.synthetic.as_ref()) {
-            (Some(output), _) => output,
-            (None, Some(output)) => output,
-            (None, None) => unreachable!("normalized exec output missing data"),
-        }
-    }
-}
-
-/// Converts a raw execution result into a uniform view that always exposes an
-/// [`ExecToolCallOutput`], synthesizing error output when the command fails
-/// before producing a response.
-pub(crate) fn normalize_exec_result(
-    result: &Result<ExecToolCallOutput, ExecError>,
-) -> NormalizedExecOutput<'_> {
-    match result {
-        Ok(output) => NormalizedExecOutput {
-            borrowed: Some(output),
-            synthetic: None,
-        },
-        Err(ExecError::Codex(CodexErr::Sandbox(SandboxErr::Timeout { output }))) => {
-            NormalizedExecOutput {
-                borrowed: Some(output.as_ref()),
-                synthetic: None,
-            }
-        }
-        Err(err) => {
-            let message = match err {
-                ExecError::Function(FunctionCallError::RespondToModel(msg)) => msg.clone(),
-                ExecError::Codex(e) => get_error_message_ui(e),
-            };
-            let synthetic = ExecToolCallOutput {
-                exit_code: -1,
-                stdout: StreamOutput::new(String::new()),
-                stderr: StreamOutput::new(message.clone()),
-                aggregated_output: StreamOutput::new(message),
-                duration: Duration::default(),
-                timed_out: false,
-            };
-            NormalizedExecOutput {
-                borrowed: None,
-                synthetic: Some(synthetic),
-            }
-        }
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use crate::error::CodexErr;
-    use crate::error::EnvVarError;
-    use crate::error::SandboxErr;
-    use crate::exec::StreamOutput;
-    use pretty_assertions::assert_eq;
-
-    fn make_output(text: &str) -> ExecToolCallOutput {
-        ExecToolCallOutput {
-            exit_code: 1,
-            stdout: StreamOutput::new(String::new()),
-            stderr: StreamOutput::new(String::new()),
-            aggregated_output: StreamOutput::new(text.to_string()),
-            duration: Duration::from_millis(123),
-            timed_out: false,
-        }
-    }
-
-    #[test]
-    fn normalize_success_borrows() {
-        let out = make_output("ok");
-        let result: Result<ExecToolCallOutput, ExecError> = Ok(out);
-        let normalized = normalize_exec_result(&result);
-        assert_eq!(normalized.event_output().aggregated_output.text, "ok");
-    }
-
-    #[test]
-    fn normalize_timeout_borrows_embedded_output() {
-        let out = make_output("timed out payload");
-        let err = CodexErr::Sandbox(SandboxErr::Timeout {
-            output: Box::new(out),
-        });
-        let result: Result<ExecToolCallOutput, ExecError> = Err(ExecError::Codex(err));
-        let normalized = normalize_exec_result(&result);
-        assert_eq!(
-            normalized.event_output().aggregated_output.text,
-            "timed out payload"
-        );
-    }
-
-    #[test]
-    fn normalize_function_error_synthesizes_payload() {
-        let err = FunctionCallError::RespondToModel("boom".to_string());
-        let result: Result<ExecToolCallOutput, ExecError> = Err(ExecError::Function(err));
-        let normalized = normalize_exec_result(&result);
-        assert_eq!(normalized.event_output().aggregated_output.text, "boom");
-    }
-
-    #[test]
-    fn normalize_codex_error_synthesizes_user_message() {
-        // Use a simple EnvVar error which formats to a clear message
-        let e = CodexErr::EnvVar(EnvVarError {
-            var: "FOO".to_string(),
-            instructions: Some("set it".to_string()),
-        });
-        let result: Result<ExecToolCallOutput, ExecError> = Err(ExecError::Codex(e));
-        let normalized = normalize_exec_result(&result);
-        assert!(
-            normalized
-                .event_output()
-                .aggregated_output
-                .text
-                .contains("Missing environment variable: `FOO`"),
-            "expected synthesized user-friendly message"
-        );
-    }
-}

codex-rs/core/src/executor/sandbox.rs

@@ -1,405 +0,0 @@
-use crate::apply_patch::ApplyPatchExec;
-use crate::codex::Session;
-use crate::exec::SandboxType;
-use crate::executor::ExecutionMode;
-use crate::executor::ExecutionRequest;
-use crate::executor::ExecutorConfig;
-use crate::executor::errors::ExecError;
-use crate::safety::SafetyCheck;
-use crate::safety::assess_command_safety;
-use crate::safety::assess_patch_safety;
-use codex_otel::otel_event_manager::OtelEventManager;
-use codex_otel::otel_event_manager::ToolDecisionSource;
-use codex_protocol::protocol::AskForApproval;
-use codex_protocol::protocol::ReviewDecision;
-use std::collections::HashSet;
-
-/// Sandbox placement options selected for an execution run, including whether
-/// to escalate after failures and whether approvals should persist.
-pub(crate) struct SandboxDecision {
-    pub(crate) initial_sandbox: SandboxType,
-    pub(crate) escalate_on_failure: bool,
-    pub(crate) record_session_approval: bool,
-}
-
-impl SandboxDecision {
-    fn auto(sandbox: SandboxType, escalate_on_failure: bool) -> Self {
-        Self {
-            initial_sandbox: sandbox,
-            escalate_on_failure,
-            record_session_approval: false,
-        }
-    }
-
-    fn user_override(record_session_approval: bool) -> Self {
-        Self {
-            initial_sandbox: SandboxType::None,
-            escalate_on_failure: false,
-            record_session_approval,
-        }
-    }
-}
-
-fn should_escalate_on_failure(approval: AskForApproval, sandbox: SandboxType) -> bool {
-    matches!(
-        (approval, sandbox),
-        (
-            AskForApproval::UnlessTrusted | AskForApproval::OnFailure,
-            SandboxType::MacosSeatbelt | SandboxType::LinuxSeccomp
-        )
-    )
-}
-
-/// Determines how a command should be sandboxed, prompting the user when
-/// policy requires explicit approval.
-#[allow(clippy::too_many_arguments)]
-pub async fn select_sandbox(
-    request: &ExecutionRequest,
-    approval_policy: AskForApproval,
-    approval_cache: HashSet<Vec<String>>,
-    config: &ExecutorConfig,
-    session: &Session,
-    sub_id: &str,
-    call_id: &str,
-    otel_event_manager: &OtelEventManager,
-) -> Result<SandboxDecision, ExecError> {
-    match &request.mode {
-        ExecutionMode::Shell => {
-            select_shell_sandbox(
-                request,
-                approval_policy,
-                approval_cache,
-                config,
-                session,
-                sub_id,
-                call_id,
-                otel_event_manager,
-            )
-            .await
-        }
-        ExecutionMode::ApplyPatch(exec) => {
-            select_apply_patch_sandbox(exec, approval_policy, config)
-        }
-    }
-}
-
-#[allow(clippy::too_many_arguments)]
-async fn select_shell_sandbox(
-    request: &ExecutionRequest,
-    approval_policy: AskForApproval,
-    approved_snapshot: HashSet<Vec<String>>,
-    config: &ExecutorConfig,
-    session: &Session,
-    sub_id: &str,
-    call_id: &str,
-    otel_event_manager: &OtelEventManager,
-) -> Result<SandboxDecision, ExecError> {
-    let command_for_safety = if request.approval_command.is_empty() {
-        request.params.command.clone()
-    } else {
-        request.approval_command.clone()
-    };
-
-    let safety = assess_command_safety(
-        &command_for_safety,
-        approval_policy,
-        &config.sandbox_policy,
-        &approved_snapshot,
-        request.params.with_escalated_permissions.unwrap_or(false),
-    );
-
-    match safety {
-        SafetyCheck::AutoApprove {
-            sandbox_type,
-            user_explicitly_approved,
-        } => {
-            let mut decision = SandboxDecision::auto(
-                sandbox_type,
-                should_escalate_on_failure(approval_policy, sandbox_type),
-            );
-            if user_explicitly_approved {
-                decision.record_session_approval = true;
-            }
-            let (decision_for_event, source) = if user_explicitly_approved {
-                (ReviewDecision::ApprovedForSession, ToolDecisionSource::User)
-            } else {
-                (ReviewDecision::Approved, ToolDecisionSource::Config)
-            };
-            otel_event_manager.tool_decision("local_shell", call_id, decision_for_event, source);
-            Ok(decision)
-        }
-        SafetyCheck::AskUser => {
-            let decision = session
-                .request_command_approval(
-                    sub_id.to_string(),
-                    call_id.to_string(),
-                    request.approval_command.clone(),
-                    request.params.cwd.clone(),
-                    request.params.justification.clone(),
-                )
-                .await;
-
-            otel_event_manager.tool_decision(
-                "local_shell",
-                call_id,
-                decision,
-                ToolDecisionSource::User,
-            );
-            match decision {
-                ReviewDecision::Approved => Ok(SandboxDecision::user_override(false)),
-                ReviewDecision::ApprovedForSession => Ok(SandboxDecision::user_override(true)),
-                ReviewDecision::Denied | ReviewDecision::Abort => {
-                    Err(ExecError::rejection("exec command rejected by user"))
-                }
-            }
-        }
-        SafetyCheck::Reject { reason } => Err(ExecError::rejection(format!(
-            "exec command rejected: {reason}"
-        ))),
-    }
-}
-
-fn select_apply_patch_sandbox(
-    exec: &ApplyPatchExec,
-    approval_policy: AskForApproval,
-    config: &ExecutorConfig,
-) -> Result<SandboxDecision, ExecError> {
-    if exec.user_explicitly_approved_this_action {
-        return Ok(SandboxDecision::user_override(false));
-    }
-
-    match assess_patch_safety(
-        &exec.action,
-        approval_policy,
-        &config.sandbox_policy,
-        &config.sandbox_cwd,
-    ) {
-        SafetyCheck::AutoApprove { sandbox_type, .. } => Ok(SandboxDecision::auto(
-            sandbox_type,
-            should_escalate_on_failure(approval_policy, sandbox_type),
-        )),
-        SafetyCheck::AskUser => Err(ExecError::rejection(
-            "patch requires approval but none was recorded",
-        )),
-        SafetyCheck::Reject { reason } => {
-            Err(ExecError::rejection(format!("patch rejected: {reason}")))
-        }
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    use super::*;
-    use crate::codex::make_session_and_context;
-    use crate::exec::ExecParams;
-    use crate::function_tool::FunctionCallError;
-    use crate::protocol::SandboxPolicy;
-    use codex_apply_patch::ApplyPatchAction;
-    use pretty_assertions::assert_eq;
-
-    #[tokio::test]
-    async fn select_apply_patch_user_override_when_explicit() {
-        let (session, ctx) = make_session_and_context();
-        let tmp = tempfile::tempdir().expect("tmp");
-        let p = tmp.path().join("a.txt");
-        let action = ApplyPatchAction::new_add_for_test(&p, "hello".to_string());
-        let exec = ApplyPatchExec {
-            action,
-            user_explicitly_approved_this_action: true,
-        };
-        let cfg = ExecutorConfig::new(SandboxPolicy::ReadOnly, std::env::temp_dir(), None);
-        let request = ExecutionRequest {
-            params: ExecParams {
-                command: vec!["apply_patch".into()],
-                cwd: std::env::temp_dir(),
-                timeout_ms: None,
-                env: std::collections::HashMap::new(),
-                with_escalated_permissions: None,
-                justification: None,
-            },
-            approval_command: vec!["apply_patch".into()],
-            mode: ExecutionMode::ApplyPatch(exec),
-            stdout_stream: None,
-            use_shell_profile: false,
-        };
-        let otel_event_manager = ctx.client.get_otel_event_manager();
-        let decision = select_sandbox(
-            &request,
-            AskForApproval::OnRequest,
-            Default::default(),
-            &cfg,
-            &session,
-            "sub",
-            "call",
-            &otel_event_manager,
-        )
-        .await
-        .expect("ok");
-        // Explicit user override runs without sandbox
-        assert_eq!(decision.initial_sandbox, SandboxType::None);
-        assert_eq!(decision.escalate_on_failure, false);
-    }
-
-    #[tokio::test]
-    async fn select_apply_patch_autoapprove_in_danger() {
-        let (session, ctx) = make_session_and_context();
-        let tmp = tempfile::tempdir().expect("tmp");
-        let p = tmp.path().join("a.txt");
-        let action = ApplyPatchAction::new_add_for_test(&p, "hello".to_string());
-        let exec = ApplyPatchExec {
-            action,
-            user_explicitly_approved_this_action: false,
-        };
-        let cfg = ExecutorConfig::new(SandboxPolicy::DangerFullAccess, std::env::temp_dir(), None);
-        let request = ExecutionRequest {
-            params: ExecParams {
-                command: vec!["apply_patch".into()],
-                cwd: std::env::temp_dir(),
-                timeout_ms: None,
-                env: std::collections::HashMap::new(),
-                with_escalated_permissions: None,
-                justification: None,
-            },
-            approval_command: vec!["apply_patch".into()],
-            mode: ExecutionMode::ApplyPatch(exec),
-            stdout_stream: None,
-            use_shell_profile: false,
-        };
-        let otel_event_manager = ctx.client.get_otel_event_manager();
-        let decision = select_sandbox(
-            &request,
-            AskForApproval::OnRequest,
-            Default::default(),
-            &cfg,
-            &session,
-            "sub",
-            "call",
-            &otel_event_manager,
-        )
-        .await
-        .expect("ok");
-        // On platforms with a sandbox, DangerFullAccess still prefers it
-        let expected = crate::safety::get_platform_sandbox().unwrap_or(SandboxType::None);
-        assert_eq!(decision.initial_sandbox, expected);
-        assert_eq!(decision.escalate_on_failure, false);
-    }
-
-    #[tokio::test]
-    async fn select_apply_patch_requires_approval_on_unless_trusted() {
-        let (session, ctx) = make_session_and_context();
-        let tempdir = tempfile::tempdir().expect("tmpdir");
-        let p = tempdir.path().join("a.txt");
-        let action = ApplyPatchAction::new_add_for_test(&p, "hello".to_string());
-        let exec = ApplyPatchExec {
-            action,
-            user_explicitly_approved_this_action: false,
-        };
-        let cfg = ExecutorConfig::new(SandboxPolicy::ReadOnly, std::env::temp_dir(), None);
-        let request = ExecutionRequest {
-            params: ExecParams {
-                command: vec!["apply_patch".into()],
-                cwd: std::env::temp_dir(),
-                timeout_ms: None,
-                env: std::collections::HashMap::new(),
-                with_escalated_permissions: None,
-                justification: None,
-            },
-            approval_command: vec!["apply_patch".into()],
-            mode: ExecutionMode::ApplyPatch(exec),
-            stdout_stream: None,
-            use_shell_profile: false,
-        };
-        let otel_event_manager = ctx.client.get_otel_event_manager();
-        let result = select_sandbox(
-            &request,
-            AskForApproval::UnlessTrusted,
-            Default::default(),
-            &cfg,
-            &session,
-            "sub",
-            "call",
-            &otel_event_manager,
-        )
-        .await;
-        match result {
-            Ok(_) => panic!("expected error"),
-            Err(ExecError::Function(FunctionCallError::RespondToModel(msg))) => {
-                assert!(msg.contains("requires approval"))
-            }
-            Err(other) => panic!("unexpected error: {other:?}"),
-        }
-    }
-
-    #[tokio::test]
-    async fn select_shell_autoapprove_in_danger_mode() {
-        let (session, ctx) = make_session_and_context();
-        let cfg = ExecutorConfig::new(SandboxPolicy::DangerFullAccess, std::env::temp_dir(), None);
-        let request = ExecutionRequest {
-            params: ExecParams {
-                command: vec!["some-unknown".into()],
-                cwd: std::env::temp_dir(),
-                timeout_ms: None,
-                env: std::collections::HashMap::new(),
-                with_escalated_permissions: None,
-                justification: None,
-            },
-            approval_command: vec!["some-unknown".into()],
-            mode: ExecutionMode::Shell,
-            stdout_stream: None,
-            use_shell_profile: false,
-        };
-        let otel_event_manager = ctx.client.get_otel_event_manager();
-        let decision = select_sandbox(
-            &request,
-            AskForApproval::OnRequest,
-            Default::default(),
-            &cfg,
-            &session,
-            "sub",
-            "call",
-            &otel_event_manager,
-        )
-        .await
-        .expect("ok");
-        assert_eq!(decision.initial_sandbox, SandboxType::None);
-        assert_eq!(decision.escalate_on_failure, false);
-    }
-
-    #[cfg(any(target_os = "macos", target_os = "linux"))]
-    #[tokio::test]
-    async fn select_shell_escalates_on_failure_with_platform_sandbox() {
-        let (session, ctx) = make_session_and_context();
-        let cfg = ExecutorConfig::new(SandboxPolicy::ReadOnly, std::env::temp_dir(), None);
-        let request = ExecutionRequest {
-            params: ExecParams {
-                // Unknown command => untrusted but not flagged dangerous
-                command: vec!["some-unknown".into()],
-                cwd: std::env::temp_dir(),
-                timeout_ms: None,
-                env: std::collections::HashMap::new(),
-                with_escalated_permissions: None,
-                justification: None,
-            },
-            approval_command: vec!["some-unknown".into()],
-            mode: ExecutionMode::Shell,
-            stdout_stream: None,
-            use_shell_profile: false,
-        };
-        let otel_event_manager = ctx.client.get_otel_event_manager();
-        let decision = select_sandbox(
-            &request,
-            AskForApproval::OnFailure,
-            Default::default(),
-            &cfg,
-            &session,
-            "sub",
-            "call",
-            &otel_event_manager,
-        )
-        .await
-        .expect("ok");
-        // On macOS/Linux we should have a platform sandbox and escalate on failure
-        assert_ne!(decision.initial_sandbox, SandboxType::None);
-        assert_eq!(decision.escalate_on_failure, true);
-    }
-}

codex-rs/core/src/features.rs

@@ -0,0 +1,258 @@
+//! Centralized feature flags and metadata.
+//!
+//! This module defines a small set of toggles that gate experimental and
+//! optional behavior across the codebase. Instead of wiring individual
+//! booleans through multiple types, call sites consult a single `Features`
+//! container attached to `Config`.
+
+use crate::config::ConfigToml;
+use crate::config_profile::ConfigProfile;
+use serde::Deserialize;
+use std::collections::BTreeMap;
+use std::collections::BTreeSet;
+
+mod legacy;
+pub(crate) use legacy::LegacyFeatureToggles;
+
+/// High-level lifecycle stage for a feature.
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum Stage {
+    Experimental,
+    Beta,
+    Stable,
+    Deprecated,
+    Removed,
+}
+
+/// Unique features toggled via configuration.
+#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub enum Feature {
+    /// Use the single unified PTY-backed exec tool.
+    UnifiedExec,
+    /// Use the streamable exec-command/write-stdin tool pair.
+    StreamableShell,
+    /// Use the official Rust MCP client (rmcp).
+    RmcpClient,
+    /// Include the plan tool.
+    PlanTool,
+    /// Include the freeform apply_patch tool.
+    ApplyPatchFreeform,
+    /// Include the view_image tool.
+    ViewImageTool,
+    /// Allow the model to request web searches.
+    WebSearchRequest,
+    /// Automatically approve all approval requests from the harness.
+    ApproveAll,
+}
+
+impl Feature {
+    pub fn key(self) -> &'static str {
+        self.info().key
+    }
+
+    pub fn stage(self) -> Stage {
+        self.info().stage
+    }
+
+    pub fn default_enabled(self) -> bool {
+        self.info().default_enabled
+    }
+
+    fn info(self) -> &'static FeatureSpec {
+        FEATURES
+            .iter()
+            .find(|spec| spec.id == self)
+            .unwrap_or_else(|| unreachable!("missing FeatureSpec for {:?}", self))
+    }
+}
+
+/// Holds the effective set of enabled features.
+#[derive(Debug, Clone, Default, PartialEq)]
+pub struct Features {
+    enabled: BTreeSet<Feature>,
+}
+
+#[derive(Debug, Clone, Default)]
+pub struct FeatureOverrides {
+    pub include_plan_tool: Option<bool>,
+    pub include_apply_patch_tool: Option<bool>,
+    pub include_view_image_tool: Option<bool>,
+    pub web_search_request: Option<bool>,
+}
+
+impl FeatureOverrides {
+    fn apply(self, features: &mut Features) {
+        LegacyFeatureToggles {
+            include_plan_tool: self.include_plan_tool,
+            include_apply_patch_tool: self.include_apply_patch_tool,
+            include_view_image_tool: self.include_view_image_tool,
+            tools_web_search: self.web_search_request,
+            ..Default::default()
+        }
+        .apply(features);
+    }
+}
+
+impl Features {
+    /// Starts with built-in defaults.
+    pub fn with_defaults() -> Self {
+        let mut set = BTreeSet::new();
+        for spec in FEATURES {
+            if spec.default_enabled {
+                set.insert(spec.id);
+            }
+        }
+        Self { enabled: set }
+    }
+
+    pub fn enabled(&self, f: Feature) -> bool {
+        self.enabled.contains(&f)
+    }
+
+    pub fn enable(&mut self, f: Feature) {
+        self.enabled.insert(f);
+    }
+
+    pub fn disable(&mut self, f: Feature) {
+        self.enabled.remove(&f);
+    }
+
+    /// Apply a table of key -> bool toggles (e.g. from TOML).
+    pub fn apply_map(&mut self, m: &BTreeMap<String, bool>) {
+        for (k, v) in m {
+            match feature_for_key(k) {
+                Some(feat) => {
+                    if *v {
+                        self.enable(feat);
+                    } else {
+                        self.disable(feat);
+                    }
+                }
+                None => {
+                    tracing::warn!("unknown feature key in config: {k}");
+                }
+            }
+        }
+    }
+
+    pub fn from_config(
+        cfg: &ConfigToml,
+        config_profile: &ConfigProfile,
+        overrides: FeatureOverrides,
+    ) -> Self {
+        let mut features = Features::with_defaults();
+
+        let base_legacy = LegacyFeatureToggles {
+            experimental_use_freeform_apply_patch: cfg.experimental_use_freeform_apply_patch,
+            experimental_use_exec_command_tool: cfg.experimental_use_exec_command_tool,
+            experimental_use_unified_exec_tool: cfg.experimental_use_unified_exec_tool,
+            experimental_use_rmcp_client: cfg.experimental_use_rmcp_client,
+            tools_web_search: cfg.tools.as_ref().and_then(|t| t.web_search),
+            tools_view_image: cfg.tools.as_ref().and_then(|t| t.view_image),
+            ..Default::default()
+        };
+        base_legacy.apply(&mut features);
+
+        if let Some(base_features) = cfg.features.as_ref() {
+            features.apply_map(&base_features.entries);
+        }
+
+        let profile_legacy = LegacyFeatureToggles {
+            include_plan_tool: config_profile.include_plan_tool,
+            include_apply_patch_tool: config_profile.include_apply_patch_tool,
+            include_view_image_tool: config_profile.include_view_image_tool,
+            experimental_use_freeform_apply_patch: config_profile
+                .experimental_use_freeform_apply_patch,
+            experimental_use_exec_command_tool: config_profile.experimental_use_exec_command_tool,
+            experimental_use_unified_exec_tool: config_profile.experimental_use_unified_exec_tool,
+            experimental_use_rmcp_client: config_profile.experimental_use_rmcp_client,
+            tools_web_search: config_profile.tools_web_search,
+            tools_view_image: config_profile.tools_view_image,
+        };
+        profile_legacy.apply(&mut features);
+        if let Some(profile_features) = config_profile.features.as_ref() {
+            features.apply_map(&profile_features.entries);
+        }
+
+        overrides.apply(&mut features);
+
+        features
+    }
+}
+
+/// Keys accepted in `[features]` tables.
+fn feature_for_key(key: &str) -> Option<Feature> {
+    for spec in FEATURES {
+        if spec.key == key {
+            return Some(spec.id);
+        }
+    }
+    legacy::feature_for_key(key)
+}
+
+/// Deserializable features table for TOML.
+#[derive(Deserialize, Debug, Clone, Default, PartialEq)]
+pub struct FeaturesToml {
+    #[serde(flatten)]
+    pub entries: BTreeMap<String, bool>,
+}
+
+/// Single, easy-to-read registry of all feature definitions.
+#[derive(Debug, Clone, Copy)]
+pub struct FeatureSpec {
+    pub id: Feature,
+    pub key: &'static str,
+    pub stage: Stage,
+    pub default_enabled: bool,
+}
+
+pub const FEATURES: &[FeatureSpec] = &[
+    FeatureSpec {
+        id: Feature::UnifiedExec,
+        key: "unified_exec",
+        stage: Stage::Experimental,
+        default_enabled: false,
+    },
+    FeatureSpec {
+        id: Feature::StreamableShell,
+        key: "streamable_shell",
+        stage: Stage::Experimental,
+        default_enabled: false,
+    },
+    FeatureSpec {
+        id: Feature::RmcpClient,
+        key: "rmcp_client",
+        stage: Stage::Experimental,
+        default_enabled: false,
+    },
+    FeatureSpec {
+        id: Feature::PlanTool,
+        key: "plan_tool",
+        stage: Stage::Stable,
+        default_enabled: false,
+    },
+    FeatureSpec {
+        id: Feature::ApplyPatchFreeform,
+        key: "apply_patch_freeform",
+        stage: Stage::Beta,
+        default_enabled: false,
+    },
+    FeatureSpec {
+        id: Feature::ViewImageTool,
+        key: "view_image_tool",
+        stage: Stage::Stable,
+        default_enabled: true,
+    },
+    FeatureSpec {
+        id: Feature::WebSearchRequest,
+        key: "web_search_request",
+        stage: Stage::Stable,
+        default_enabled: false,
+    },
+    FeatureSpec {
+        id: Feature::ApproveAll,
+        key: "approve_all",
+        stage: Stage::Experimental,
+        default_enabled: false,
+    },
+];

codex-rs/core/src/features/legacy.rs

@@ -0,0 +1,158 @@
+use super::Feature;
+use super::Features;
+use tracing::info;
+
+#[derive(Clone, Copy)]
+struct Alias {
+    legacy_key: &'static str,
+    feature: Feature,
+}
+
+const ALIASES: &[Alias] = &[
+    Alias {
+        legacy_key: "experimental_use_unified_exec_tool",
+        feature: Feature::UnifiedExec,
+    },
+    Alias {
+        legacy_key: "experimental_use_exec_command_tool",
+        feature: Feature::StreamableShell,
+    },
+    Alias {
+        legacy_key: "experimental_use_rmcp_client",
+        feature: Feature::RmcpClient,
+    },
+    Alias {
+        legacy_key: "experimental_use_freeform_apply_patch",
+        feature: Feature::ApplyPatchFreeform,
+    },
+    Alias {
+        legacy_key: "include_apply_patch_tool",
+        feature: Feature::ApplyPatchFreeform,
+    },
+    Alias {
+        legacy_key: "include_plan_tool",
+        feature: Feature::PlanTool,
+    },
+    Alias {
+        legacy_key: "include_view_image_tool",
+        feature: Feature::ViewImageTool,
+    },
+    Alias {
+        legacy_key: "web_search",
+        feature: Feature::WebSearchRequest,
+    },
+];
+
+pub(crate) fn feature_for_key(key: &str) -> Option<Feature> {
+    ALIASES
+        .iter()
+        .find(|alias| alias.legacy_key == key)
+        .map(|alias| {
+            log_alias(alias.legacy_key, alias.feature);
+            alias.feature
+        })
+}
+
+#[derive(Debug, Default)]
+pub struct LegacyFeatureToggles {
+    pub include_plan_tool: Option<bool>,
+    pub include_apply_patch_tool: Option<bool>,
+    pub include_view_image_tool: Option<bool>,
+    pub experimental_use_freeform_apply_patch: Option<bool>,
+    pub experimental_use_exec_command_tool: Option<bool>,
+    pub experimental_use_unified_exec_tool: Option<bool>,
+    pub experimental_use_rmcp_client: Option<bool>,
+    pub tools_web_search: Option<bool>,
+    pub tools_view_image: Option<bool>,
+}
+
+impl LegacyFeatureToggles {
+    pub fn apply(self, features: &mut Features) {
+        set_if_some(
+            features,
+            Feature::PlanTool,
+            self.include_plan_tool,
+            "include_plan_tool",
+        );
+        set_if_some(
+            features,
+            Feature::ApplyPatchFreeform,
+            self.include_apply_patch_tool,
+            "include_apply_patch_tool",
+        );
+        set_if_some(
+            features,
+            Feature::ApplyPatchFreeform,
+            self.experimental_use_freeform_apply_patch,
+            "experimental_use_freeform_apply_patch",
+        );
+        set_if_some(
+            features,
+            Feature::StreamableShell,
+            self.experimental_use_exec_command_tool,
+            "experimental_use_exec_command_tool",
+        );
+        set_if_some(
+            features,
+            Feature::UnifiedExec,
+            self.experimental_use_unified_exec_tool,
+            "experimental_use_unified_exec_tool",
+        );
+        set_if_some(
+            features,
+            Feature::RmcpClient,
+            self.experimental_use_rmcp_client,
+            "experimental_use_rmcp_client",
+        );
+        set_if_some(
+            features,
+            Feature::WebSearchRequest,
+            self.tools_web_search,
+            "tools.web_search",
+        );
+        set_if_some(
+            features,
+            Feature::ViewImageTool,
+            self.include_view_image_tool,
+            "include_view_image_tool",
+        );
+        set_if_some(
+            features,
+            Feature::ViewImageTool,
+            self.tools_view_image,
+            "tools.view_image",
+        );
+    }
+}
+
+fn set_if_some(
+    features: &mut Features,
+    feature: Feature,
+    maybe_value: Option<bool>,
+    alias_key: &'static str,
+) {
+    if let Some(enabled) = maybe_value {
+        set_feature(features, feature, enabled);
+        log_alias(alias_key, feature);
+    }
+}
+
+fn set_feature(features: &mut Features, feature: Feature, enabled: bool) {
+    if enabled {
+        features.enable(feature);
+    } else {
+        features.disable(feature);
+    }
+}
+
+fn log_alias(alias: &str, feature: Feature) {
+    let canonical = feature.key();
+    if alias == canonical {
+        return;
+    }
+    info!(
+        %alias,
+        canonical,
+        "legacy feature toggle detected; prefer `[features].{canonical}`"
+    );
+}

codex-rs/core/src/function_tool.rs

@@ -4,4 +4,11 @@ use thiserror::Error;
 pub enum FunctionCallError {
     #[error("{0}")]
     RespondToModel(String),
+    #[error("{0}")]
+    #[allow(dead_code)] // TODO(jif) fix in a follow-up PR
+    Denied(String),
+    #[error("LocalShellCall without call_id or id")]
+    MissingLocalShellCallId,
+    #[error("Fatal error: {0}")]
+    Fatal(String),
 }

codex-rs/core/src/landlock.rs

@@ -40,7 +40,7 @@ where
 }
 
 /// Converts the sandbox policy into the CLI invocation for `codex-linux-sandbox`.
-fn create_linux_sandbox_command_args(
+pub(crate) fn create_linux_sandbox_command_args(
     command: Vec<String>,
     sandbox_policy: &SandboxPolicy,
     sandbox_policy_cwd: &Path,
@@ -56,7 +56,9 @@ fn create_linux_sandbox_command_args(
         serde_json::to_string(sandbox_policy).expect("Failed to serialize SandboxPolicy to JSON");
 
     let mut linux_cmd: Vec<String> = vec![
+        "--sandbox-policy-cwd".to_string(),
         sandbox_policy_cwd,
+        "--sandbox-policy".to_string(),
         sandbox_policy_json,
         // Separator so that command arguments starting with `-` are not parsed as
         // options of the helper itself.

codex-rs/core/src/lib.rs

@@ -13,11 +13,11 @@ mod client;
 mod client_common;
 pub mod codex;
 mod codex_conversation;
-pub mod token_data;
 pub use codex_conversation::CodexConversation;
 mod command_safety;
 pub mod config;
 pub mod config_edit;
+pub mod config_loader;
 pub mod config_profile;
 pub mod config_types;
 mod conversation_history;
@@ -25,17 +25,19 @@ pub mod custom_prompts;
 mod environment_context;
 pub mod error;
 pub mod exec;
-mod exec_command;
 pub mod exec_env;
-pub mod executor;
+pub mod features;
 mod flags;
 pub mod git_info;
 pub mod landlock;
+pub mod mcp;
 mod mcp_connection_manager;
 mod mcp_tool_call;
 mod message_history;
 mod model_provider_info;
 pub mod parse_command;
+pub mod sandboxing;
+pub mod token_data;
 mod truncate;
 mod unified_exec;
 mod user_instructions;
@@ -56,8 +58,6 @@ pub use auth::CodexAuth;
 pub mod default_client;
 pub mod model_family;
 mod openai_model_info;
-mod openai_tools;
-pub mod plan_tool;
 pub mod project_doc;
 mod rollout;
 pub(crate) mod safety;
@@ -65,9 +65,10 @@ pub mod seatbelt;
 pub mod shell;
 pub mod spawn;
 pub mod terminal;
-mod tool_apply_patch;
+mod tools;
 pub mod turn_diff_tracker;
 pub use rollout::ARCHIVED_SESSIONS_SUBDIR;
+pub use rollout::INTERACTIVE_SESSION_SOURCES;
 pub use rollout::RolloutRecorder;
 pub use rollout::SESSIONS_SUBDIR;
 pub use rollout::SessionMeta;
@@ -104,5 +105,4 @@ pub use codex_protocol::models::LocalShellExecAction;
 pub use codex_protocol::models::LocalShellStatus;
 pub use codex_protocol::models::ReasoningItemContent;
 pub use codex_protocol::models::ResponseItem;
-
 pub mod otel_init;

codex-rs/core/src/mcp/auth.rs

@@ -0,0 +1,62 @@
+use std::collections::HashMap;
+
+use anyhow::Result;
+use codex_protocol::protocol::McpAuthStatus;
+use codex_rmcp_client::OAuthCredentialsStoreMode;
+use codex_rmcp_client::determine_streamable_http_auth_status;
+use futures::future::join_all;
+use tracing::warn;
+
+use crate::config_types::McpServerConfig;
+use crate::config_types::McpServerTransportConfig;
+
+pub async fn compute_auth_statuses<'a, I>(
+    servers: I,
+    store_mode: OAuthCredentialsStoreMode,
+) -> HashMap<String, McpAuthStatus>
+where
+    I: IntoIterator<Item = (&'a String, &'a McpServerConfig)>,
+{
+    let futures = servers.into_iter().map(|(name, config)| {
+        let name = name.clone();
+        let config = config.clone();
+        async move {
+            let status = match compute_auth_status(&name, &config, store_mode).await {
+                Ok(status) => status,
+                Err(error) => {
+                    warn!("failed to determine auth status for MCP server `{name}`: {error:?}");
+                    McpAuthStatus::Unsupported
+                }
+            };
+            (name, status)
+        }
+    });
+
+    join_all(futures).await.into_iter().collect()
+}
+
+async fn compute_auth_status(
+    server_name: &str,
+    config: &McpServerConfig,
+    store_mode: OAuthCredentialsStoreMode,
+) -> Result<McpAuthStatus> {
+    match &config.transport {
+        McpServerTransportConfig::Stdio { .. } => Ok(McpAuthStatus::Unsupported),
+        McpServerTransportConfig::StreamableHttp {
+            url,
+            bearer_token_env_var,
+            http_headers,
+            env_http_headers,
+        } => {
+            determine_streamable_http_auth_status(
+                server_name,
+                url,
+                bearer_token_env_var.as_deref(),
+                http_headers.clone(),
+                env_http_headers.clone(),
+                store_mode,
+            )
+            .await
+        }
+    }
+}

codex-rs/core/src/mcp/mod.rs

@@ -0,0 +1 @@
+pub mod auth;

codex-rs/core/src/mcp_connection_manager.rs

@@ -8,7 +8,9 @@
 
 use std::collections::HashMap;
 use std::collections::HashSet;
+use std::env;
 use std::ffi::OsString;
+use std::path::PathBuf;
 use std::sync::Arc;
 use std::time::Duration;
 
@@ -16,9 +18,18 @@ use anyhow::Context;
 use anyhow::Result;
 use anyhow::anyhow;
 use codex_mcp_client::McpClient;
+use codex_rmcp_client::OAuthCredentialsStoreMode;
 use codex_rmcp_client::RmcpClient;
 use mcp_types::ClientCapabilities;
 use mcp_types::Implementation;
+use mcp_types::ListResourceTemplatesRequestParams;
+use mcp_types::ListResourceTemplatesResult;
+use mcp_types::ListResourcesRequestParams;
+use mcp_types::ListResourcesResult;
+use mcp_types::ReadResourceRequestParams;
+use mcp_types::ReadResourceResult;
+use mcp_types::Resource;
+use mcp_types::ResourceTemplate;
 use mcp_types::Tool;
 
 use serde_json::json;
@@ -54,8 +65,8 @@ fn qualify_tools(tools: Vec<ToolInfo>) -> HashMap<String, ToolInfo> {
     let mut qualified_tools = HashMap::new();
     for tool in tools {
         let mut qualified_name = format!(
-            "{}{}{}",
-            tool.server_name, MCP_TOOL_NAME_DELIMITER, tool.tool_name
+            "mcp{}{}{}{}",
+            MCP_TOOL_NAME_DELIMITER, tool.server_name, MCP_TOOL_NAME_DELIMITER, tool.tool_name
         );
         if qualified_name.len() > MAX_TOOL_NAME_LENGTH {
             let mut hasher = Sha1::new();
@@ -100,35 +111,52 @@ enum McpClientAdapter {
 }
 
 impl McpClientAdapter {
+    #[allow(clippy::too_many_arguments)]
     async fn new_stdio_client(
         use_rmcp_client: bool,
         program: OsString,
         args: Vec<OsString>,
         env: Option<HashMap<String, String>>,
+        env_vars: Vec<String>,
+        cwd: Option<PathBuf>,
         params: mcp_types::InitializeRequestParams,
         startup_timeout: Duration,
     ) -> Result<Self> {
-        info!(
-            "new_stdio_client use_rmcp_client: {use_rmcp_client} program: {program:?} args: {args:?} env: {env:?} params: {params:?} startup_timeout: {startup_timeout:?}"
-        );
         if use_rmcp_client {
-            let client = Arc::new(RmcpClient::new_stdio_client(program, args, env).await?);
+            let client =
+                Arc::new(RmcpClient::new_stdio_client(program, args, env, &env_vars, cwd).await?);
             client.initialize(params, Some(startup_timeout)).await?;
             Ok(McpClientAdapter::Rmcp(client))
         } else {
-            let client = Arc::new(McpClient::new_stdio_client(program, args, env).await?);
+            let client =
+                Arc::new(McpClient::new_stdio_client(program, args, env, &env_vars, cwd).await?);
             client.initialize(params, Some(startup_timeout)).await?;
             Ok(McpClientAdapter::Legacy(client))
         }
     }
 
+    #[allow(clippy::too_many_arguments)]
     async fn new_streamable_http_client(
+        server_name: String,
         url: String,
         bearer_token: Option<String>,
+        http_headers: Option<HashMap<String, String>>,
+        env_http_headers: Option<HashMap<String, String>>,
         params: mcp_types::InitializeRequestParams,
         startup_timeout: Duration,
+        store_mode: OAuthCredentialsStoreMode,
     ) -> Result<Self> {
-        let client = Arc::new(RmcpClient::new_streamable_http_client(url, bearer_token)?);
+        let client = Arc::new(
+            RmcpClient::new_streamable_http_client(
+                &server_name,
+                &url,
+                bearer_token,
+                http_headers,
+                env_http_headers,
+                store_mode,
+            )
+            .await?,
+        );
         client.initialize(params, Some(startup_timeout)).await?;
         Ok(McpClientAdapter::Rmcp(client))
     }
@@ -144,6 +172,47 @@ impl McpClientAdapter {
         }
     }
 
+    async fn list_resources(
+        &self,
+        params: Option<mcp_types::ListResourcesRequestParams>,
+        timeout: Option<Duration>,
+    ) -> Result<mcp_types::ListResourcesResult> {
+        match self {
+            McpClientAdapter::Legacy(_) => Ok(ListResourcesResult {
+                next_cursor: None,
+                resources: Vec::new(),
+            }),
+            McpClientAdapter::Rmcp(client) => client.list_resources(params, timeout).await,
+        }
+    }
+
+    async fn read_resource(
+        &self,
+        params: mcp_types::ReadResourceRequestParams,
+        timeout: Option<Duration>,
+    ) -> Result<mcp_types::ReadResourceResult> {
+        match self {
+            McpClientAdapter::Legacy(_) => Err(anyhow!(
+                "resources/read is not supported by legacy MCP clients"
+            )),
+            McpClientAdapter::Rmcp(client) => client.read_resource(params, timeout).await,
+        }
+    }
+
+    async fn list_resource_templates(
+        &self,
+        params: Option<mcp_types::ListResourceTemplatesRequestParams>,
+        timeout: Option<Duration>,
+    ) -> Result<mcp_types::ListResourceTemplatesResult> {
+        match self {
+            McpClientAdapter::Legacy(_) => Ok(ListResourceTemplatesResult {
+                next_cursor: None,
+                resource_templates: Vec::new(),
+            }),
+            McpClientAdapter::Rmcp(client) => client.list_resource_templates(params, timeout).await,
+        }
+    }
+
     async fn call_tool(
         &self,
         name: String,
@@ -182,6 +251,7 @@ impl McpConnectionManager {
     pub async fn new(
         mcp_servers: HashMap<String, McpServerConfig>,
         use_rmcp_client: bool,
+        store_mode: OAuthCredentialsStoreMode,
     ) -> Result<(Self, ClientStartErrors)> {
         // Early exit if no servers are configured.
         if mcp_servers.is_empty() {
@@ -202,22 +272,21 @@ impl McpConnectionManager {
                 continue;
             }
 
-            if matches!(
-                cfg.transport,
-                McpServerTransportConfig::StreamableHttp { .. }
-            ) && !use_rmcp_client
-            {
-                info!(
-                    "skipping MCP server `{}` configured with url because rmcp client is disabled",
-                    server_name
-                );
+            if !cfg.enabled {
                 continue;
             }
 
             let startup_timeout = cfg.startup_timeout_sec.unwrap_or(DEFAULT_STARTUP_TIMEOUT);
             let tool_timeout = cfg.tool_timeout_sec.unwrap_or(DEFAULT_TOOL_TIMEOUT);
 
-            let use_rmcp_client_flag = use_rmcp_client;
+            let resolved_bearer_token = match &cfg.transport {
+                McpServerTransportConfig::StreamableHttp {
+                    bearer_token_env_var,
+                    ..
+                } => resolve_bearer_token(&server_name, bearer_token_env_var.as_deref()),
+                _ => Ok(None),
+            };
+
             join_set.spawn(async move {
                 let McpServerConfig { transport, .. } = cfg;
                 let params = mcp_types::InitializeRequestParams {
@@ -242,25 +311,42 @@ impl McpConnectionManager {
                 };
 
                 let client = match transport {
-                    McpServerTransportConfig::Stdio { command, args, env } => {
+                    McpServerTransportConfig::Stdio {
+                        command,
+                        args,
+                        env,
+                        env_vars,
+                        cwd,
+                    } => {
                         let command_os: OsString = command.into();
                         let args_os: Vec<OsString> = args.into_iter().map(Into::into).collect();
                         McpClientAdapter::new_stdio_client(
-                            use_rmcp_client_flag,
+                            use_rmcp_client,
                             command_os,
                             args_os,
                             env,
-                            params.clone(),
+                            env_vars,
+                            cwd,
+                            params,
                             startup_timeout,
                         )
                         .await
                     }
-                    McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
+                    McpServerTransportConfig::StreamableHttp {
+                        url,
+                        http_headers,
+                        env_http_headers,
+                        ..
+                    } => {
                         McpClientAdapter::new_streamable_http_client(
+                            server_name.clone(),
                             url,
-                            bearer_token,
+                            resolved_bearer_token.unwrap_or_default(),
+                            http_headers,
+                            env_http_headers,
                             params,
                             startup_timeout,
+                            store_mode,
                         )
                         .await
                     }
@@ -312,7 +398,7 @@ impl McpConnectionManager {
         Ok((Self { clients, tools }, errors))
     }
 
-    /// Returns a single map that contains **all** tools. Each key is the
+    /// Returns a single map that contains all tools. Each key is the
     /// fully-qualified name for the tool.
     pub fn list_all_tools(&self) -> HashMap<String, Tool> {
         self.tools
@@ -321,6 +407,133 @@ impl McpConnectionManager {
             .collect()
     }
 
+    /// Returns a single map that contains all resources. Each key is the
+    /// server name and the value is a vector of resources.
+    pub async fn list_all_resources(&self) -> HashMap<String, Vec<Resource>> {
+        let mut join_set = JoinSet::new();
+
+        for (server_name, managed_client) in &self.clients {
+            let server_name_cloned = server_name.clone();
+            let client_clone = managed_client.client.clone();
+            let timeout = managed_client.tool_timeout;
+
+            join_set.spawn(async move {
+                let mut collected: Vec<Resource> = Vec::new();
+                let mut cursor: Option<String> = None;
+
+                loop {
+                    let params = cursor.as_ref().map(|next| ListResourcesRequestParams {
+                        cursor: Some(next.clone()),
+                    });
+                    let response = match client_clone.list_resources(params, timeout).await {
+                        Ok(result) => result,
+                        Err(err) => return (server_name_cloned, Err(err)),
+                    };
+
+                    collected.extend(response.resources);
+
+                    match response.next_cursor {
+                        Some(next) => {
+                            if cursor.as_ref() == Some(&next) {
+                                return (
+                                    server_name_cloned,
+                                    Err(anyhow!("resources/list returned duplicate cursor")),
+                                );
+                            }
+                            cursor = Some(next);
+                        }
+                        None => return (server_name_cloned, Ok(collected)),
+                    }
+                }
+            });
+        }
+
+        let mut aggregated: HashMap<String, Vec<Resource>> = HashMap::new();
+
+        while let Some(join_res) = join_set.join_next().await {
+            match join_res {
+                Ok((server_name, Ok(resources))) => {
+                    aggregated.insert(server_name, resources);
+                }
+                Ok((server_name, Err(err))) => {
+                    warn!("Failed to list resources for MCP server '{server_name}': {err:#}");
+                }
+                Err(err) => {
+                    warn!("Task panic when listing resources for MCP server: {err:#}");
+                }
+            }
+        }
+
+        aggregated
+    }
+
+    /// Returns a single map that contains all resource templates. Each key is the
+    /// server name and the value is a vector of resource templates.
+    pub async fn list_all_resource_templates(&self) -> HashMap<String, Vec<ResourceTemplate>> {
+        let mut join_set = JoinSet::new();
+
+        for (server_name, managed_client) in &self.clients {
+            let server_name_cloned = server_name.clone();
+            let client_clone = managed_client.client.clone();
+            let timeout = managed_client.tool_timeout;
+
+            join_set.spawn(async move {
+                let mut collected: Vec<ResourceTemplate> = Vec::new();
+                let mut cursor: Option<String> = None;
+
+                loop {
+                    let params = cursor
+                        .as_ref()
+                        .map(|next| ListResourceTemplatesRequestParams {
+                            cursor: Some(next.clone()),
+                        });
+                    let response = match client_clone.list_resource_templates(params, timeout).await
+                    {
+                        Ok(result) => result,
+                        Err(err) => return (server_name_cloned, Err(err)),
+                    };
+
+                    collected.extend(response.resource_templates);
+
+                    match response.next_cursor {
+                        Some(next) => {
+                            if cursor.as_ref() == Some(&next) {
+                                return (
+                                    server_name_cloned,
+                                    Err(anyhow!(
+                                        "resources/templates/list returned duplicate cursor"
+                                    )),
+                                );
+                            }
+                            cursor = Some(next);
+                        }
+                        None => return (server_name_cloned, Ok(collected)),
+                    }
+                }
+            });
+        }
+
+        let mut aggregated: HashMap<String, Vec<ResourceTemplate>> = HashMap::new();
+
+        while let Some(join_res) = join_set.join_next().await {
+            match join_res {
+                Ok((server_name, Ok(templates))) => {
+                    aggregated.insert(server_name, templates);
+                }
+                Ok((server_name, Err(err))) => {
+                    warn!(
+                        "Failed to list resource templates for MCP server '{server_name}': {err:#}"
+                    );
+                }
+                Err(err) => {
+                    warn!("Task panic when listing resource templates for MCP server: {err:#}");
+                }
+            }
+        }
+
+        aggregated
+    }
+
     /// Invoke the tool indicated by the (server, tool) pair.
     pub async fn call_tool(
         &self,
@@ -332,7 +545,7 @@ impl McpConnectionManager {
             .clients
             .get(server)
             .ok_or_else(|| anyhow!("unknown MCP server '{server}'"))?;
-        let client = managed.client.clone();
+        let client = &managed.client;
         let timeout = managed.tool_timeout;
 
         client
@@ -341,6 +554,64 @@ impl McpConnectionManager {
             .with_context(|| format!("tool call failed for `{server}/{tool}`"))
     }
 
+    /// List resources from the specified server.
+    pub async fn list_resources(
+        &self,
+        server: &str,
+        params: Option<ListResourcesRequestParams>,
+    ) -> Result<ListResourcesResult> {
+        let managed = self
+            .clients
+            .get(server)
+            .ok_or_else(|| anyhow!("unknown MCP server '{server}'"))?;
+        let client = managed.client.clone();
+        let timeout = managed.tool_timeout;
+
+        client
+            .list_resources(params, timeout)
+            .await
+            .with_context(|| format!("resources/list failed for `{server}`"))
+    }
+
+    /// List resource templates from the specified server.
+    pub async fn list_resource_templates(
+        &self,
+        server: &str,
+        params: Option<ListResourceTemplatesRequestParams>,
+    ) -> Result<ListResourceTemplatesResult> {
+        let managed = self
+            .clients
+            .get(server)
+            .ok_or_else(|| anyhow!("unknown MCP server '{server}'"))?;
+        let client = managed.client.clone();
+        let timeout = managed.tool_timeout;
+
+        client
+            .list_resource_templates(params, timeout)
+            .await
+            .with_context(|| format!("resources/templates/list failed for `{server}`"))
+    }
+
+    /// Read a resource from the specified server.
+    pub async fn read_resource(
+        &self,
+        server: &str,
+        params: ReadResourceRequestParams,
+    ) -> Result<ReadResourceResult> {
+        let managed = self
+            .clients
+            .get(server)
+            .ok_or_else(|| anyhow!("unknown MCP server '{server}'"))?;
+        let client = managed.client.clone();
+        let timeout = managed.tool_timeout;
+        let uri = params.uri.clone();
+
+        client
+            .read_resource(params, timeout)
+            .await
+            .with_context(|| format!("resources/read failed for `{server}` ({uri})"))
+    }
+
     pub fn parse_tool_name(&self, tool_name: &str) -> Option<(String, String)> {
         self.tools
             .get(tool_name)
@@ -348,8 +619,35 @@ impl McpConnectionManager {
     }
 }
 
+fn resolve_bearer_token(
+    server_name: &str,
+    bearer_token_env_var: Option<&str>,
+) -> Result<Option<String>> {
+    let Some(env_var) = bearer_token_env_var else {
+        return Ok(None);
+    };
+
+    match env::var(env_var) {
+        Ok(value) => {
+            if value.is_empty() {
+                Err(anyhow!(
+                    "Environment variable {env_var} for MCP server '{server_name}' is empty"
+                ))
+            } else {
+                Ok(Some(value))
+            }
+        }
+        Err(env::VarError::NotPresent) => Err(anyhow!(
+            "Environment variable {env_var} for MCP server '{server_name}' is not set"
+        )),
+        Err(env::VarError::NotUnicode(_)) => Err(anyhow!(
+            "Environment variable {env_var} for MCP server '{server_name}' contains invalid Unicode"
+        )),
+    }
+}
+
 /// Query every server for its available tools and return a single map that
-/// contains **all** tools. Each key is the fully-qualified name for the tool.
+/// contains all tools. Each key is the fully-qualified name for the tool.
 async fn list_all_tools(clients: &HashMap<String, ManagedClient>) -> Result<Vec<ToolInfo>> {
     let mut join_set = JoinSet::new();
 
@@ -443,8 +741,8 @@ mod tests {
         let qualified_tools = qualify_tools(tools);
 
         assert_eq!(qualified_tools.len(), 2);
-        assert!(qualified_tools.contains_key("server1__tool1"));
-        assert!(qualified_tools.contains_key("server1__tool2"));
+        assert!(qualified_tools.contains_key("mcp__server1__tool1"));
+        assert!(qualified_tools.contains_key("mcp__server1__tool2"));
     }
 
     #[test]
@@ -458,7 +756,7 @@ mod tests {
 
         // Only the first tool should remain, the second is skipped
         assert_eq!(qualified_tools.len(), 1);
-        assert!(qualified_tools.contains_key("server1__duplicate_tool"));
+        assert!(qualified_tools.contains_key("mcp__server1__duplicate_tool"));
     }
 
     #[test]
@@ -486,13 +784,13 @@ mod tests {
         assert_eq!(keys[0].len(), 64);
         assert_eq!(
             keys[0],
-            "my_server__extremely_lena02e507efc5a9de88637e436690364fd4219e4ef"
+            "mcp__my_server__extremel119a2b97664e41363932dc84de21e2ff1b93b3e9"
         );
 
         assert_eq!(keys[1].len(), 64);
         assert_eq!(
             keys[1],
-            "my_server__yet_another_e1c3987bd9c50b826cbe1687966f79f0c602d19ca"
+            "mcp__my_server__yet_anot419a82a89325c1b477274a41f8c65ea5f3a7f341"
         );
     }
 }