Merge branch 'main' into patch-guard

1. If Codex detects that a `codex mcp add -url …` server supports oauth,
it will auto-initiate the login flow.
2. If the TUI starts and a MCP server supports oauth but isn't logged
in, it will give the user an explicit warning telling them to log in.

.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/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/workflows/issue-deduplicator.yml

@@ -105,6 +105,10 @@ jobs:
 
             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) {
@@ -113,7 +117,8 @@ jobs:
             }
 
             const lines = [
-              'Potential duplicates detected:',
+              '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)*'];

.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 +94,118 @@ 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
+
       - name: Stage artifacts
         shell: bash
         run: |
@@ -157,6 +264,29 @@ 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 }}

AGENTS.md

@@ -73,3 +73,28 @@ If you don’t have the tool:
 ### 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

@@ -61,7 +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
 

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/Cargo.lock

@@ -300,6 +300,12 @@ dependencies = [
  "wait-timeout",
 ]
 
+[[package]]
+name = "assert_matches"
+version = "1.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "9b34d609dfbaf33d6889b2b7106d3ca345eacad44200913df5ba02bfd31d2ba9"
+
 [[package]]
 name = "async-broadcast"
 version = "0.7.2"
@@ -871,6 +877,7 @@ version = "0.0.0"
 dependencies = [
  "anyhow",
  "assert_cmd",
+ "assert_matches",
  "pretty_assertions",
  "similar",
  "tempfile",
@@ -933,6 +940,7 @@ version = "0.0.0"
 dependencies = [
  "anyhow",
  "assert_cmd",
+ "assert_matches",
  "clap",
  "clap_complete",
  "codex-app-server",
@@ -980,12 +988,11 @@ dependencies = [
  "reqwest",
  "serde",
  "serde_json",
- "throbber-widgets-tui",
  "tokio",
  "tokio-stream",
  "tracing",
  "tracing-subscriber",
- "unicode-width 0.1.14",
+ "unicode-width 0.2.1",
 ]
 
 [[package]]
@@ -1022,6 +1029,7 @@ dependencies = [
  "anyhow",
  "askama",
  "assert_cmd",
+ "assert_matches",
  "async-channel",
  "async-trait",
  "base64",
@@ -1043,6 +1051,7 @@ dependencies = [
  "escargot",
  "eventsource-stream",
  "futures",
+ "ignore",
  "indexmap 2.10.0",
  "landlock",
  "libc",
@@ -1061,6 +1070,7 @@ dependencies = [
  "serde_json",
  "serial_test",
  "sha1",
+ "sha2",
  "shlex",
  "similar",
  "strum_macros 0.27.2",
@@ -1162,6 +1172,7 @@ dependencies = [
 name = "codex-git-tooling"
 version = "0.0.0"
 dependencies = [
+ "assert_matches",
  "pretty_assertions",
  "tempfile",
  "thiserror 2.0.16",
@@ -1249,6 +1260,7 @@ dependencies = [
 name = "codex-ollama"
 version = "0.0.0"
 dependencies = [
+ "assert_matches",
  "async-stream",
  "bytes",
  "codex-core",
@@ -1342,6 +1354,7 @@ version = "0.0.0"
 dependencies = [
  "anyhow",
  "axum",
+ "codex-protocol",
  "dirs",
  "futures",
  "keyring",
@@ -1367,6 +1380,7 @@ version = "0.0.0"
 dependencies = [
  "anyhow",
  "arboard",
+ "assert_matches",
  "async-stream",
  "base64",
  "chrono",
@@ -1413,6 +1427,8 @@ dependencies = [
  "tracing",
  "tracing-appender",
  "tracing-subscriber",
+ "tree-sitter-bash",
+ "tree-sitter-highlight",
  "unicode-segmentation",
  "unicode-width 0.2.1",
  "url",
@@ -1432,6 +1448,7 @@ dependencies = [
 name = "codex-utils-readiness"
 version = "0.0.0"
 dependencies = [
+ "assert_matches",
  "async-trait",
  "thiserror 2.0.16",
  "time",
@@ -1562,9 +1579,12 @@ dependencies = [
  "anyhow",
  "assert_cmd",
  "codex-core",
+ "notify",
+ "regex-lite",
  "serde_json",
  "tempfile",
  "tokio",
+ "walkdir",
  "wiremock",
 ]
 
@@ -2355,6 +2375,15 @@ dependencies = [
  "percent-encoding",
 ]
 
+[[package]]
+name = "fsevent-sys"
+version = "4.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "76ee7a02da4d231650c7cea31349b889be2f45ddb3ef3032d2ec8185f6313fd2"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "futures"
 version = "0.3.31"
@@ -3041,6 +3070,26 @@ version = "2.0.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "f4c7245a08504955605670dbf141fceab975f15ca21570696aebe9d2e71576bd"
 
+[[package]]
+name = "inotify"
+version = "0.11.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f37dccff2791ab604f9babef0ba14fbe0be30bd368dc541e2b08d07c8aa908f3"
+dependencies = [
+ "bitflags 2.9.1",
+ "inotify-sys",
+ "libc",
+]
+
+[[package]]
+name = "inotify-sys"
+version = "0.1.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e05c02b5e89bff3b946cedeca278abc628fe811e604f027c45a8aa3cf793d0eb"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "inout"
 version = "0.1.4"
@@ -3241,6 +3290,26 @@ dependencies = [
  "zeroize",
 ]
 
+[[package]]
+name = "kqueue"
+version = "1.1.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "eac30106d7dce88daf4a3fcb4879ea939476d5074a9b7ddd0fb97fa4bed5596a"
+dependencies = [
+ "kqueue-sys",
+ "libc",
+]
+
+[[package]]
+name = "kqueue-sys"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "ed9625ffda8729b85e45cf04090035ac368927b8cebc34898e7c120f52e4838b"
+dependencies = [
+ "bitflags 1.3.2",
+ "libc",
+]
+
 [[package]]
 name = "lalrpop"
 version = "0.19.12"
@@ -3640,6 +3709,30 @@ version = "0.3.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "61807f77802ff30975e01f4f071c8ba10c022052f98b3294119f3e615d13e5be"
 
+[[package]]
+name = "notify"
+version = "8.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4d3d07927151ff8575b7087f245456e549fea62edf0ec4e565a5ee50c8402bc3"
+dependencies = [
+ "bitflags 2.9.1",
+ "fsevent-sys",
+ "inotify",
+ "kqueue",
+ "libc",
+ "log",
+ "mio",
+ "notify-types",
+ "walkdir",
+ "windows-sys 0.60.2",
+]
+
+[[package]]
+name = "notify-types"
+version = "2.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5e0826a989adedc2a244799e823aece04662b66609d96af8dff7ac6df9a8925d"
+
 [[package]]
 name = "nu-ansi-term"
 version = "0.50.1"
@@ -4757,8 +4850,9 @@ dependencies = [
 
 [[package]]
 name = "rmcp"
-version = "0.7.0"
-source = "git+https://github.com/modelcontextprotocol/rust-sdk?rev=c0b777c7f784ba2d456b03c2ec3b98c9b28b5e10#c0b777c7f784ba2d456b03c2ec3b98c9b28b5e10"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6f35acda8f89fca5fd8c96cae3c6d5b4c38ea0072df4c8030915f3b5ff469c1c"
 dependencies = [
  "base64",
  "bytes",
@@ -4790,8 +4884,9 @@ dependencies = [
 
 [[package]]
 name = "rmcp-macros"
-version = "0.7.0"
-source = "git+https://github.com/modelcontextprotocol/rust-sdk?rev=c0b777c7f784ba2d456b03c2ec3b98c9b28b5e10#c0b777c7f784ba2d456b03c2ec3b98c9b28b5e10"
+version = "0.8.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c9f1d5220aaa23b79c3d02e18f7a554403b3ccea544bbb6c69d6bcb3e854a274"
 dependencies = [
  "darling 0.21.3",
  "proc-macro2",
@@ -5827,16 +5922,6 @@ dependencies = [
  "cfg-if",
 ]
 
-[[package]]
-name = "throbber-widgets-tui"
-version = "0.8.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1d36b5738d666a2b4c91b7c24998a8588db724b3107258343ebf8824bf55b06d"
-dependencies = [
- "rand 0.8.5",
- "ratatui",
-]
-
 [[package]]
 name = "tiff"
 version = "0.10.3"
@@ -6014,6 +6099,7 @@ dependencies = [
  "bytes",
  "futures-core",
  "futures-sink",
+ "futures-util",
  "pin-project-lite",
  "tokio",
 ]
@@ -6255,9 +6341,9 @@ dependencies = [
 
 [[package]]
 name = "tree-sitter"
-version = "0.25.9"
+version = "0.25.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ccd2a058a86cfece0bf96f7cce1021efef9c8ed0e892ab74639173e5ed7a34fa"
+checksum = "78f873475d258561b06f1c595d93308a7ed124d9977cb26b148c2084a4a3cc87"
 dependencies = [
  "cc",
  "regex",
@@ -6277,6 +6363,18 @@ dependencies = [
  "tree-sitter-language",
 ]
 
+[[package]]
+name = "tree-sitter-highlight"
+version = "0.25.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "adc5f880ad8d8f94e88cb81c3557024cf1a8b75e3b504c50481ed4f5a6006ff3"
+dependencies = [
+ "regex",
+ "streaming-iterator",
+ "thiserror 2.0.16",
+ "tree-sitter",
+]
+
 [[package]]
 name = "tree-sitter-language"
 version = "0.1.5"

codex-rs/Cargo.toml

@@ -83,6 +83,7 @@ ansi-to-tui = "7.0.0"
 anyhow = "1"
 arboard = "3"
 askama = "0.12"
+assert_matches = "1.5.0"
 assert_cmd = "2"
 async-channel = "2.3.1"
 async-stream = "0.3.6"
@@ -121,6 +122,7 @@ 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"
@@ -142,8 +144,7 @@ rand = "0.9"
 ratatui = "0.29.0"
 regex-lite = "0.1.7"
 reqwest = "0.12"
-# When this gets bumped, remove the patch below.
-rmcp = { version = "0.7.0", default-features = false }
+rmcp = { version = "0.8.0", default-features = false }
 schemars = "0.8.22"
 seccompiler = "0.5.0"
 serde = "1"
@@ -175,8 +176,9 @@ 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"
 unicode-segmentation = "1.12.0"
 unicode-width = "0.2"
@@ -250,5 +252,3 @@ ratatui = { git = "https://github.com/nornagon/ratatui", branch = "nornagon-v0.2
 
 # Uncomment to debug local changes.
 # rmcp = { path = "../../rust-sdk/crates/rmcp" }
-# 0.7.0 + https://github.com/modelcontextprotocol/rust-sdk/pull/459 which includes dynamic .well-known discovery which is required for Figma.
-rmcp = { git = "https://github.com/modelcontextprotocol/rust-sdk", rev = "c0b777c7f784ba2d456b03c2ec3b98c9b28b5e10", default-features = false }

codex-rs/README.md

@@ -23,9 +23,15 @@ 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
@@ -71,9 +77,13 @@ To test to see what happens when a command is run under the sandbox provided by
 
 ```
 # 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/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/cli/Cargo.toml

@@ -47,6 +47,7 @@ tokio = { workspace = true, features = [
 ] }
 
 [dev-dependencies]
+assert_matches = { workspace = true }
 assert_cmd = { workspace = true }
 predicates = { workspace = true }
 pretty_assertions = { workspace = true }

codex-rs/cli/src/main.rs

@@ -26,6 +26,8 @@ use supports_color::Stream;
 mod mcp_cmd;
 
 use crate::mcp_cmd::McpCli;
+use codex_core::config::Config;
+use codex_core::config::ConfigOverrides;
 
 /// Codex CLI
 ///
@@ -45,6 +47,9 @@ struct MultitoolCli {
     #[clap(flatten)]
     pub config_overrides: CliConfigOverrides,
 
+    #[clap(flatten)]
+    pub feature_toggles: FeatureToggles,
+
     #[clap(flatten)]
     interactive: TuiCli,
 
@@ -76,8 +81,9 @@ 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")]
@@ -96,6 +102,9 @@ enum Subcommand {
     /// Internal: run the responses API proxy.
     #[clap(hide = true)]
     ResponsesApiProxy(ResponsesApiProxyArgs),
+
+    /// Inspect feature flags.
+    Features(FeaturesCli),
 }
 
 #[derive(Debug, Parser)]
@@ -121,18 +130,20 @@ 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)]
@@ -154,9 +165,7 @@ struct LoginCommand {
     )]
     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)
@@ -230,6 +239,53 @@ fn print_exit_messages(exit_info: AppExitInfo) {
     }
 }
 
+#[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
 /// skip this in debug builds to avoid interfering with debugging.
 #[ctor::ctor]
@@ -247,11 +303,17 @@ 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(
@@ -291,7 +353,8 @@ 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?;
+            print_exit_messages(exit_info);
         }
         Some(Subcommand::Login(mut login_cli)) => {
             prepend_config_flags(
@@ -341,8 +404,8 @@ 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(),
@@ -353,7 +416,7 @@ 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(),
@@ -379,6 +442,30 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
         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(())
@@ -472,6 +559,7 @@ 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;
 
@@ -481,6 +569,7 @@ mod tests {
             interactive,
             config_overrides: root_overrides,
             subcommand,
+            feature_toggles: _,
         } = cli;
 
         let Subcommand::Resume(ResumeCommand {
@@ -604,14 +693,14 @@ 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

@@ -4,6 +4,7 @@ use anyhow::Context;
 use anyhow::Result;
 use anyhow::anyhow;
 use anyhow::bail;
+use clap::ArgGroup;
 use codex_common::CliConfigOverrides;
 use codex_core::config::Config;
 use codex_core::config::ConfigOverrides;
@@ -12,8 +13,12 @@ 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.
 ///
@@ -77,13 +82,61 @@ 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)]
@@ -138,39 +191,61 @@ impl McpCli {
 
 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,
+            }
+        }
+        AddMcpTransportArgs {
+            streamable_http:
+                Some(AddMcpStreamableHttpArgs {
+                    url,
+                    bearer_token_env_var,
+                }),
+            ..
+        } => McpServerTransportConfig::StreamableHttp {
+            url,
+            bearer_token_env_var,
         },
+        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,
     };
@@ -182,6 +257,17 @@ async fn run_add(config_overrides: &CliConfigOverrides, add_args: AddArgs) -> Re
 
     println!("Added global MCP server '{name}'.");
 
+    if let McpServerTransportConfig::StreamableHttp {
+        url,
+        bearer_token_env_var: None,
+    } = 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).await?;
+        println!("Successfully logged in.");
+    }
+
     Ok(())
 }
 
@@ -219,7 +305,7 @@ async fn run_login(config_overrides: &CliConfigOverrides, login_args: LoginArgs)
         .await
         .context("failed to load configuration")?;
 
-    if !config.use_experimental_use_rmcp_client {
+    if !config.features.enabled(Feature::RmcpClient) {
         bail!(
             "OAuth login is only supported when experimental_use_rmcp_client is true in config.toml."
         );
@@ -236,7 +322,7 @@ async fn run_login(config_overrides: &CliConfigOverrides, login_args: LoginArgs)
         _ => bail!("OAuth login is only supported for streamable HTTP servers."),
     };
 
-    perform_oauth_login(&name, &url).await?;
+    perform_oauth_login(&name, &url, config.mcp_oauth_credentials_store_mode).await?;
     println!("Successfully logged in to MCP server '{name}'.");
     Ok(())
 }
@@ -259,7 +345,7 @@ async fn run_logout(config_overrides: &CliConfigOverrides, logout_args: LogoutAr
         _ => bail!("OAuth logout is only supported for streamable_http transports."),
     };
 
-    match delete_oauth_tokens(&name, &url) {
+    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}")),
@@ -276,11 +362,20 @@ async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) ->
 
     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!({
                         "type": "stdio",
@@ -288,17 +383,21 @@ async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) ->
                         "args": args,
                         "env": env,
                     }),
-                    McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
+                    McpServerTransportConfig::StreamableHttp {
+                        url,
+                        bearer_token_env_var,
+                    } => {
                         serde_json::json!({
                             "type": "streamable_http",
                             "url": url,
-                            "bearer_token": bearer_token,
+                            "bearer_token_env_var": bearer_token_env_var,
                         })
                     }
                 };
 
                 serde_json::json!({
                     "name": name,
+                    "enabled": cfg.enabled,
                     "transport": transport,
                     "startup_timeout_sec": cfg
                         .startup_timeout_sec
@@ -306,6 +405,7 @@ async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) ->
                     "tool_timeout_sec": cfg
                         .tool_timeout_sec
                         .map(|timeout| timeout.as_secs_f64()),
+                    "auth_status": auth_status,
                 })
             })
             .collect();
@@ -319,8 +419,8 @@ async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) ->
         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; 6]> = Vec::new();
+    let mut http_rows: Vec<[String; 5]> = Vec::new();
 
     for (name, cfg) in entries {
         match &cfg.transport {
@@ -343,21 +443,59 @@ async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) ->
                             .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 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,
+                    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(),
+            "Status".len(),
+            "Auth".len(),
+        ];
         for row in &stdio_rows {
             for (i, cell) in row.iter().enumerate() {
                 widths[i] = widths[i].max(cell.len());
@@ -365,28 +503,36 @@ async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) ->
         }
 
         println!(
-            "{:<name_w$}  {:<cmd_w$}  {:<args_w$}  {:<env_w$}",
-            "Name",
-            "Command",
-            "Args",
-            "Env",
+            "{name:<name_w$}  {command:<cmd_w$}  {args:<args_w$}  {env:<env_w$}  {status:<status_w$}  {auth:<auth_w$}",
+            name = "Name",
+            command = "Command",
+            args = "Args",
+            env = "Env",
+            status = "Status",
+            auth = "Auth",
             name_w = widths[0],
             cmd_w = widths[1],
             args_w = widths[2],
             env_w = widths[3],
+            status_w = widths[4],
+            auth_w = widths[5],
         );
 
         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$}  {env:<env_w$}  {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(),
+                status = row[4].as_str(),
+                auth = row[5].as_str(),
                 name_w = widths[0],
                 cmd_w = widths[1],
                 args_w = widths[2],
                 env_w = widths[3],
+                status_w = widths[4],
+                auth_w = widths[5],
             );
         }
     }
@@ -396,7 +542,13 @@ async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) ->
     }
 
     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());
@@ -404,24 +556,32 @@ async fn run_list(config_overrides: &CliConfigOverrides, list_args: ListArgs) ->
         }
 
         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],
             );
         }
     }
@@ -447,14 +607,18 @@ async fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Re
                 "args": args,
                 "env": env,
             }),
-            McpServerTransportConfig::StreamableHttp { url, bearer_token } => serde_json::json!({
+            McpServerTransportConfig::StreamableHttp {
+                url,
+                bearer_token_env_var,
+            } => serde_json::json!({
                 "type": "streamable_http",
                 "url": url,
-                "bearer_token": bearer_token,
+                "bearer_token_env_var": bearer_token_env_var,
             }),
         };
         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
@@ -468,6 +632,7 @@ async fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Re
     }
 
     println!("{}", get_args.name);
+    println!("  enabled: {}", server.enabled);
     match &server.transport {
         McpServerTransportConfig::Stdio { command, args, env } => {
             println!("  transport: stdio");
@@ -493,11 +658,14 @@ async fn run_get(config_overrides: &CliConfigOverrides, get_args: GetArgs) -> Re
             };
             println!("  env: {env_display}");
         }
-        McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
+        McpServerTransportConfig::StreamableHttp {
+            url,
+            bearer_token_env_var,
+        } => {
             println!("  transport: streamable_http");
             println!("  url: {url}");
-            let bearer = bearer_token.as_deref().unwrap_or("-");
-            println!("  bearer_token: {bearer}");
+            let env_var = bearer_token_env_var.as_deref().unwrap_or("-");
+            println!("  bearer_token_env_var: {env_var}");
         }
     }
     if let Some(timeout) = server.startup_timeout_sec {

codex-rs/cli/tests/mcp_add_remove.rs

@@ -35,6 +35,7 @@ async fn add_and_remove_server_updates_global_config() -> Result<()> {
         }
         other => panic!("unexpected transport: {other:?}"),
     }
+    assert!(docs.enabled);
 
     let mut remove_cmd = codex_command(codex_home.path())?;
     remove_cmd
@@ -90,6 +91,122 @@ async 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,
+        } => {
+            assert_eq!(url, "https://example.com/mcp");
+            assert!(bearer_token_env_var.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,
+        } => {
+            assert_eq!(url, "https://example.com/issues");
+            assert_eq!(bearer_token_env_var.as_deref(), Some("GITHUB_TOKEN"));
+        }
+        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 +1,7 @@
 use std::path::Path;
 
 use anyhow::Result;
+use predicates::prelude::PredicateBooleanExt;
 use predicates::str::contains;
 use pretty_assertions::assert_eq;
 use serde_json::Value as JsonValue;
@@ -53,6 +54,10 @@ 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("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 +69,7 @@ fn list_and_get_render_expected_output() -> Result<()> {
         json!([
           {
             "name": "docs",
+            "enabled": true,
             "transport": {
               "type": "stdio",
               "command": "docs-server",
@@ -76,7 +82,8 @@ fn list_and_get_render_expected_output() -> Result<()> {
               }
             },
             "startup_timeout_sec": null,
-            "tool_timeout_sec": null
+            "tool_timeout_sec": null,
+            "auth_status": "unsupported"
           }
         ]
         )
@@ -91,6 +98,7 @@ 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("enabled: true"));
     assert!(stdout.contains("remove: codex mcp remove docs"));
 
     let mut get_json_cmd = codex_command(codex_home.path())?;
@@ -98,7 +106,7 @@ 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/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(|home| codex_login::AuthManager::new(home, false))
-            .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

@@ -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/core/Cargo.toml

@@ -33,6 +33,7 @@ env-flags = { workspace = true }
 eventsource-stream = { workspace = true }
 futures = { workspace = true }
 indexmap = { workspace = true }
+ignore = { workspace = true }
 libc = { workspace = true }
 mcp-types = { workspace = true }
 os_info = { workspace = true }
@@ -43,6 +44,7 @@ reqwest = { workspace = true, features = ["json", "stream"] }
 serde = { workspace = true, features = ["derive"] }
 serde_json = { workspace = true }
 sha1 = { workspace = true }
+sha2 = { workspace = true }
 shlex = { workspace = true }
 similar = { workspace = true }
 strum_macros = { workspace = true }
@@ -61,7 +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"] }
@@ -89,6 +91,7 @@ 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 }

codex-rs/core/README.md

@@ -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/src/chat_completions.rs

@@ -389,10 +389,12 @@ 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

@@ -47,6 +47,7 @@ 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::util::backoff;
 use codex_otel::otel_event_manager::OtelEventManager;
@@ -123,8 +124,16 @@ 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(
@@ -165,7 +174,11 @@ 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");
@@ -244,7 +257,7 @@ 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) => {
@@ -272,6 +285,7 @@ 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());
@@ -294,6 +308,7 @@ 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()
@@ -649,10 +664,12 @@ 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:#}");
@@ -932,6 +949,7 @@ fn is_context_window_error(error: &Error) -> bool {
 #[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;
@@ -1391,10 +1409,7 @@ mod tests {
         let resp: ErrorResponse =
             serde_json::from_str(json).expect("should deserialize old 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\"");
@@ -1409,7 +1424,7 @@ mod tests {
         let resp: ErrorResponse =
             serde_json::from_str(json).expect("should deserialize old 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

@@ -159,12 +159,16 @@ fn build_structured_output(parsed: &ExecOutputJson) -> String {
         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(parsed.output.clone());
+    sections.push(output);
 
     sections.join("\n")
 }
@@ -177,6 +181,13 @@ fn extract_total_output_lines(output: &str) -> Option<u32> {
     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,

codex-rs/core/src/codebase_change_notice.rs

@@ -0,0 +1,168 @@
+use std::fmt::Write;
+
+use codex_protocol::models::ContentItem;
+use codex_protocol::models::ResponseItem;
+
+use crate::codebase_snapshot::SnapshotDiff;
+
+pub(crate) const CODEBASE_CHANGE_NOTICE_MAX_PATHS: usize = 40;
+
+#[derive(Clone, Debug, PartialEq, Eq)]
+pub(crate) struct CodebaseChangeNotice {
+    added: Vec<String>,
+    removed: Vec<String>,
+    modified: Vec<String>,
+    truncated: bool,
+}
+
+impl CodebaseChangeNotice {
+    pub(crate) fn new(diff: SnapshotDiff, limit: usize) -> Self {
+        let mut remaining = limit;
+        let mut truncated = false;
+
+        let added = take_paths(diff.added, &mut remaining, &mut truncated);
+        let removed = take_paths(diff.removed, &mut remaining, &mut truncated);
+        let modified = take_paths(diff.modified, &mut remaining, &mut truncated);
+
+        Self {
+            added,
+            removed,
+            modified,
+            truncated,
+        }
+    }
+
+    pub(crate) fn is_empty(&self) -> bool {
+        self.added.is_empty() && self.removed.is_empty() && self.modified.is_empty()
+    }
+
+    pub(crate) fn serialize_to_xml(&self) -> String {
+        let mut output = String::new();
+        if self.truncated {
+            let _ = writeln!(output, "<codebase_changes truncated=\"true\">");
+        } else {
+            let _ = writeln!(output, "<codebase_changes>");
+        }
+
+        let mut summary_parts = Vec::new();
+        if !self.added.is_empty() {
+            summary_parts.push(format!("added {}", self.added.len()));
+        }
+        if !self.removed.is_empty() {
+            summary_parts.push(format!("removed {}", self.removed.len()));
+        }
+        if !self.modified.is_empty() {
+            summary_parts.push(format!("modified {}", self.modified.len()));
+        }
+
+        if summary_parts.is_empty() {
+            let _ = writeln!(output, "  <summary>no changes</summary>");
+        } else {
+            let summary = summary_parts.join(", ");
+            let _ = writeln!(output, "  <summary>{summary}</summary>");
+        }
+
+        serialize_section(&mut output, "added", &self.added);
+        serialize_section(&mut output, "removed", &self.removed);
+        serialize_section(&mut output, "modified", &self.modified);
+        if self.truncated {
+            let _ = writeln!(output, "  <note>additional paths omitted</note>");
+        }
+
+        let _ = writeln!(output, "</codebase_changes>");
+        output
+    }
+}
+
+fn take_paths(mut paths: Vec<String>, remaining: &mut usize, truncated: &mut bool) -> Vec<String> {
+    if *remaining == 0 {
+        if !paths.is_empty() {
+            *truncated = true;
+        }
+        return Vec::new();
+    }
+
+    if paths.len() > *remaining {
+        paths.truncate(*remaining);
+        *truncated = true;
+    }
+
+    *remaining -= paths.len();
+    paths
+}
+
+fn serialize_section(output: &mut String, tag: &str, paths: &[String]) {
+    if paths.is_empty() {
+        return;
+    }
+
+    let _ = writeln!(output, "  <{tag}>");
+    for path in paths {
+        let _ = writeln!(output, "    <path>{}</path>", escape_xml(path));
+    }
+    let _ = writeln!(output, "  </{tag}>");
+}
+
+fn escape_xml(value: &str) -> String {
+    let mut escaped = String::with_capacity(value.len());
+    for ch in value.chars() {
+        match ch {
+            '&' => escaped.push_str("&amp;"),
+            '<' => escaped.push_str("&lt;"),
+            '>' => escaped.push_str("&gt;"),
+            '"' => escaped.push_str("&quot;"),
+            '\'' => escaped.push_str("&apos;"),
+            other => escaped.push(other),
+        }
+    }
+    escaped
+}
+
+impl From<CodebaseChangeNotice> for ResponseItem {
+    fn from(notice: CodebaseChangeNotice) -> Self {
+        ResponseItem::Message {
+            id: None,
+            role: "user".to_string(),
+            content: vec![ContentItem::InputText {
+                text: notice.serialize_to_xml(),
+            }],
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use pretty_assertions::assert_eq;
+
+    #[test]
+    fn constructs_notice_with_limit() {
+        let diff = SnapshotDiff {
+            added: vec!["a.rs".to_string(), "b.rs".to_string()],
+            removed: vec!["c.rs".to_string()],
+            modified: vec!["d.rs".to_string(), "e.rs".to_string()],
+        };
+
+        let notice = CodebaseChangeNotice::new(diff, 3);
+        assert!(notice.truncated);
+        assert_eq!(
+            notice.added.len() + notice.removed.len() + notice.modified.len(),
+            3
+        );
+    }
+
+    #[test]
+    fn serializes_notice() {
+        let diff = SnapshotDiff {
+            added: vec!["src/lib.rs".to_string()],
+            removed: Vec::new(),
+            modified: vec!["src/main.rs".to_string()],
+        };
+        let notice = CodebaseChangeNotice::new(diff, CODEBASE_CHANGE_NOTICE_MAX_PATHS);
+        let xml = notice.serialize_to_xml();
+        assert!(xml.contains("<added>"));
+        assert!(xml.contains("<modified>"));
+        assert!(xml.contains("src/lib.rs"));
+        assert!(xml.contains("src/main.rs"));
+    }
+}

codex-rs/core/src/codebase_snapshot.rs

@@ -0,0 +1,278 @@
+use std::borrow::Cow;
+use std::collections::BTreeMap;
+use std::fs::File;
+use std::io::Read;
+use std::path::Path;
+use std::path::PathBuf;
+use std::time::SystemTime;
+
+use anyhow::Context;
+use anyhow::Result;
+use ignore::WalkBuilder;
+use sha2::Digest;
+use sha2::Sha256;
+use tokio::task;
+use tracing::warn;
+
+#[derive(Clone, Debug, PartialEq, Eq)]
+pub(crate) struct CodebaseSnapshot {
+    root: PathBuf,
+    entries: BTreeMap<String, EntryFingerprint>,
+    root_digest: DigestBytes,
+}
+
+#[derive(Clone, Debug, PartialEq, Eq)]
+pub(crate) struct EntryFingerprint {
+    pub kind: EntryKind,
+    pub digest: DigestBytes,
+    pub size: u64,
+    pub modified_millis: Option<u128>,
+}
+
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+#[repr(u8)]
+pub(crate) enum EntryKind {
+    File,
+    Symlink,
+}
+
+#[derive(Clone, Debug, PartialEq, Eq, Default)]
+pub(crate) struct SnapshotDiff {
+    pub added: Vec<String>,
+    pub removed: Vec<String>,
+    pub modified: Vec<String>,
+}
+
+impl SnapshotDiff {
+    pub fn is_empty(&self) -> bool {
+        self.added.is_empty() && self.removed.is_empty() && self.modified.is_empty()
+    }
+}
+
+pub(crate) type DigestBytes = [u8; 32];
+
+impl CodebaseSnapshot {
+    pub(crate) async fn capture(root: PathBuf) -> Result<Self> {
+        task::spawn_blocking(move || Self::from_disk(&root))
+            .await
+            .map_err(|e| anyhow::anyhow!("codebase snapshot task failed: {e}"))?
+    }
+
+    pub(crate) fn from_disk(root: &Path) -> Result<Self> {
+        if !root.exists() {
+            return Ok(Self::empty(root));
+        }
+
+        let mut entries: BTreeMap<String, EntryFingerprint> = BTreeMap::new();
+
+        let mut walker = WalkBuilder::new(root);
+        walker
+            .hidden(false)
+            .git_ignore(true)
+            .git_exclude(true)
+            .parents(true)
+            .ignore(true)
+            .follow_links(false);
+
+        for result in walker.build() {
+            let entry = match result {
+                Ok(entry) => entry,
+                Err(err) => {
+                    warn!("codebase snapshot failed to read entry: {err}");
+                    continue;
+                }
+            };
+
+            let path = entry.path();
+            if entry.depth() == 0 {
+                continue;
+            }
+
+            let relative = match path.strip_prefix(root) {
+                Ok(rel) => rel,
+                Err(_) => continue,
+            };
+            if relative.as_os_str().is_empty() {
+                continue;
+            }
+            let rel_string = normalize_rel_path(relative);
+
+            let file_type = match entry.file_type() {
+                Some(file_type) => file_type,
+                None => continue,
+            };
+
+            if file_type.is_dir() {
+                continue;
+            }
+
+            if file_type.is_file() {
+                match fingerprint_file(path) {
+                    Ok(fp) => {
+                        entries.insert(rel_string, fp);
+                    }
+                    Err(err) => {
+                        warn!(
+                            "codebase snapshot failed to hash file {}: {err}",
+                            path.display()
+                        );
+                    }
+                }
+                continue;
+            }
+
+            if file_type.is_symlink() {
+                match fingerprint_symlink(path) {
+                    Ok(fp) => {
+                        entries.insert(rel_string, fp);
+                    }
+                    Err(err) => {
+                        warn!(
+                            "codebase snapshot failed to hash symlink {}: {err}",
+                            path.display()
+                        );
+                    }
+                }
+                continue;
+            }
+        }
+
+        let root_digest = compute_root_digest(&entries);
+
+        Ok(Self {
+            root: root.to_path_buf(),
+            entries,
+            root_digest,
+        })
+    }
+
+    pub(crate) fn diff(&self, newer: &CodebaseSnapshot) -> SnapshotDiff {
+        let mut diff = SnapshotDiff::default();
+
+        for (path, fingerprint) in &newer.entries {
+            match self.entries.get(path) {
+                None => diff.added.push(path.clone()),
+                Some(existing) if existing != fingerprint => diff.modified.push(path.clone()),
+                _ => {}
+            }
+        }
+
+        for path in self.entries.keys() {
+            if !newer.entries.contains_key(path) {
+                diff.removed.push(path.clone());
+            }
+        }
+
+        diff
+    }
+
+    pub(crate) fn root(&self) -> &Path {
+        &self.root
+    }
+
+    fn empty(root: &Path) -> Self {
+        Self {
+            root: root.to_path_buf(),
+            entries: BTreeMap::new(),
+            root_digest: Sha256::digest(b"").into(),
+        }
+    }
+}
+
+fn fingerprint_file(path: &Path) -> Result<EntryFingerprint> {
+    let metadata = path
+        .metadata()
+        .with_context(|| format!("metadata {}", path.display()))?;
+    let mut file = File::open(path).with_context(|| format!("open {}", path.display()))?;
+
+    let mut hasher = Sha256::new();
+    let mut buf = [0u8; 64 * 1024];
+    loop {
+        let read = file.read(&mut buf)?;
+        if read == 0 {
+            break;
+        }
+        hasher.update(&buf[..read]);
+    }
+
+    Ok(EntryFingerprint {
+        kind: EntryKind::File,
+        digest: hasher.finalize().into(),
+        size: metadata.len(),
+        modified_millis: metadata.modified().ok().and_then(system_time_to_millis),
+    })
+}
+
+fn fingerprint_symlink(path: &Path) -> Result<EntryFingerprint> {
+    let target =
+        std::fs::read_link(path).with_context(|| format!("read_link {}", path.display()))?;
+    let mut hasher = Sha256::new();
+    let target_str = normalize_rel_path(&target);
+    hasher.update(target_str.as_bytes());
+    Ok(EntryFingerprint {
+        kind: EntryKind::Symlink,
+        digest: hasher.finalize().into(),
+        size: 0,
+        modified_millis: None,
+    })
+}
+
+fn compute_root_digest(entries: &BTreeMap<String, EntryFingerprint>) -> DigestBytes {
+    let mut hasher = Sha256::new();
+    for (path, fingerprint) in entries {
+        hasher.update(path.as_bytes());
+        hasher.update(fingerprint.digest);
+        hasher.update([fingerprint.kind as u8]);
+        hasher.update(fingerprint.size.to_le_bytes());
+        if let Some(modified) = fingerprint.modified_millis {
+            hasher.update(modified.to_le_bytes());
+        }
+    }
+    hasher.finalize().into()
+}
+
+fn normalize_rel_path(path: &Path) -> String {
+    let s = path_to_cow(path);
+    if s.is_empty() {
+        String::new()
+    } else {
+        s.replace('\\', "/")
+    }
+}
+
+fn path_to_cow(path: &Path) -> Cow<'_, str> {
+    path.to_string_lossy()
+}
+
+fn system_time_to_millis(ts: SystemTime) -> Option<u128> {
+    ts.duration_since(SystemTime::UNIX_EPOCH)
+        .map(|duration| duration.as_millis())
+        .ok()
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use pretty_assertions::assert_eq;
+    use tempfile::tempdir;
+
+    #[test]
+    fn diff_tracks_added_modified_removed() {
+        let dir = tempdir().unwrap();
+        let root = dir.path();
+
+        std::fs::write(root.join("file_a.txt"), "alpha").unwrap();
+        std::fs::write(root.join("file_b.txt"), "bravo").unwrap();
+        let snapshot_one = CodebaseSnapshot::from_disk(root).unwrap();
+
+        std::fs::write(root.join("file_a.txt"), "alpha-updated").unwrap();
+        std::fs::remove_file(root.join("file_b.txt")).unwrap();
+        std::fs::write(root.join("file_c.txt"), "charlie").unwrap();
+        let snapshot_two = CodebaseSnapshot::from_disk(root).unwrap();
+
+        let diff = snapshot_one.diff(&snapshot_two);
+        assert_eq!(diff.added, vec!["file_c.txt".to_string()]);
+        assert_eq!(diff.modified, vec!["file_a.txt".to_string()]);
+        assert_eq!(diff.removed, vec!["file_b.txt".to_string()]);
+    }
+}

codex-rs/core/src/codex.rs

@@ -1,5 +1,6 @@
 use std::borrow::Cow;
 use std::fmt::Debug;
+use std::path::Path;
 use std::path::PathBuf;
 use std::sync::Arc;
 use std::sync::atomic::AtomicU64;
@@ -17,13 +18,16 @@ use codex_apply_patch::ApplyPatchAction;
 use codex_protocol::ConversationId;
 use codex_protocol::protocol::ConversationPathResponseEvent;
 use codex_protocol::protocol::ExitedReviewModeEvent;
+use codex_protocol::protocol::McpAuthStatus;
 use codex_protocol::protocol::ReviewRequest;
 use codex_protocol::protocol::RolloutItem;
 use codex_protocol::protocol::SessionSource;
 use codex_protocol::protocol::TaskStartedEvent;
 use codex_protocol::protocol::TurnAbortReason;
 use codex_protocol::protocol::TurnContextItem;
+use futures::future::BoxFuture;
 use futures::prelude::*;
+use futures::stream::FuturesOrdered;
 use mcp_types::CallToolResult;
 use serde_json;
 use serde_json::Value;
@@ -40,6 +44,9 @@ use crate::apply_patch::convert_apply_patch_to_protocol;
 use crate::client::ModelClient;
 use crate::client_common::Prompt;
 use crate::client_common::ResponseEvent;
+use crate::codebase_change_notice::CODEBASE_CHANGE_NOTICE_MAX_PATHS;
+use crate::codebase_change_notice::CodebaseChangeNotice;
+use crate::codebase_snapshot::CodebaseSnapshot;
 use crate::config::Config;
 use crate::config_types::ShellEnvironmentPolicy;
 use crate::conversation_history::ConversationHistory;
@@ -55,6 +62,7 @@ use crate::exec_command::WriteStdinParams;
 use crate::executor::Executor;
 use crate::executor::ExecutorConfig;
 use crate::executor::normalize_exec_result;
+use crate::mcp::auth::compute_auth_statuses;
 use crate::mcp_connection_manager::McpConnectionManager;
 use crate::model_family::find_family_for_model;
 use crate::openai_model_info::get_model_info;
@@ -96,6 +104,7 @@ use crate::rollout::RolloutRecorderParams;
 use crate::shell;
 use crate::state::ActiveTurn;
 use crate::state::SessionServices;
+use crate::state::TaskKind;
 use crate::tasks::CompactTask;
 use crate::tasks::RegularTask;
 use crate::tasks::ReviewTask;
@@ -361,14 +370,32 @@ impl Session {
 
         let mcp_fut = McpConnectionManager::new(
             config.mcp_servers.clone(),
-            config.use_experimental_use_rmcp_client,
+            config
+                .features
+                .enabled(crate::features::Feature::RmcpClient),
+            config.mcp_oauth_credentials_store_mode,
         );
         let default_shell_fut = shell::default_user_shell();
         let history_meta_fut = crate::message_history::history_metadata(&config);
+        let auth_statuses_fut = compute_auth_statuses(
+            config.mcp_servers.iter(),
+            config.mcp_oauth_credentials_store_mode,
+        );
 
         // Join all independent futures.
-        let (rollout_recorder, mcp_res, default_shell, (history_log_id, history_entry_count)) =
-            tokio::join!(rollout_fut, mcp_fut, default_shell_fut, history_meta_fut);
+        let (
+            rollout_recorder,
+            mcp_res,
+            default_shell,
+            (history_log_id, history_entry_count),
+            auth_statuses,
+        ) = tokio::join!(
+            rollout_fut,
+            mcp_fut,
+            default_shell_fut,
+            history_meta_fut,
+            auth_statuses_fut
+        );
 
         let rollout_recorder = rollout_recorder.map_err(|e| {
             error!("failed to initialize rollout recorder: {e:#}");
@@ -395,11 +422,24 @@ impl Session {
         // Surface individual client start-up failures to the user.
         if !failed_clients.is_empty() {
             for (server_name, err) in failed_clients {
-                let message = format!("MCP client for `{server_name}` failed to start: {err:#}");
-                error!("{message}");
+                let log_message =
+                    format!("MCP client for `{server_name}` failed to start: {err:#}");
+                error!("{log_message}");
+                let display_message = if matches!(
+                    auth_statuses.get(&server_name),
+                    Some(McpAuthStatus::NotLoggedIn)
+                ) {
+                    format!(
+                        "The {server_name} MCP server is not logged in. Run `codex mcp login {server_name}` to log in."
+                    )
+                } else {
+                    log_message
+                };
                 post_session_configured_error_events.push(Event {
                     id: INITIAL_SUBMIT_ID.to_owned(),
-                    msg: EventMsg::Error(ErrorEvent { message }),
+                    msg: EventMsg::Error(ErrorEvent {
+                        message: display_message,
+                    }),
                 });
             }
         }
@@ -442,12 +482,7 @@ impl Session {
             client,
             tools_config: ToolsConfig::new(&ToolsConfigParams {
                 model_family: &config.model_family,
-                include_plan_tool: config.include_plan_tool,
-                include_apply_patch_tool: config.include_apply_patch_tool,
-                include_web_search_request: config.tools_web_search_request,
-                use_streamable_shell_tool: config.use_experimental_streamable_shell_tool,
-                include_view_image_tool: config.include_view_image_tool,
-                experimental_unified_exec_tool: config.use_experimental_unified_exec_tool,
+                features: &config.features,
             }),
             user_instructions,
             base_instructions,
@@ -715,6 +750,73 @@ impl Session {
         self.persist_rollout_items(&rollout_items).await;
     }
 
+    async fn stored_snapshot_for_root(&self, root: &Path) -> Option<CodebaseSnapshot> {
+        let state = self.state.lock().await;
+        state
+            .codebase_snapshot
+            .as_ref()
+            .filter(|snapshot| snapshot.root() == root)
+            .cloned()
+    }
+
+    async fn set_codebase_snapshot(&self, snapshot: CodebaseSnapshot) {
+        let mut state = self.state.lock().await;
+        state.codebase_snapshot = Some(snapshot);
+    }
+
+    pub(crate) async fn emit_codebase_delta_if_changed(
+        &self,
+        turn_context: &TurnContext,
+        sub_id: &str,
+    ) -> anyhow::Result<()> {
+        let cwd = turn_context.cwd.clone();
+        let previous = self.stored_snapshot_for_root(&cwd).await;
+        let latest = CodebaseSnapshot::capture(cwd.clone()).await?;
+
+        if let Some(previous_snapshot) = previous {
+            let diff = previous_snapshot.diff(&latest);
+            if diff.is_empty() {
+                self.set_codebase_snapshot(latest).await;
+                return Ok(());
+            }
+
+            let notice = CodebaseChangeNotice::new(diff, CODEBASE_CHANGE_NOTICE_MAX_PATHS);
+            if notice.is_empty() {
+                self.set_codebase_snapshot(latest).await;
+                return Ok(());
+            }
+
+            let response_item: ResponseItem = notice.into();
+            self.record_conversation_items(std::slice::from_ref(&response_item))
+                .await;
+
+            for msg in
+                map_response_item_to_event_messages(&response_item, self.show_raw_agent_reasoning())
+            {
+                let event = Event {
+                    id: sub_id.to_string(),
+                    msg,
+                };
+                self.send_event(event).await;
+            }
+
+            self.set_codebase_snapshot(latest).await;
+            return Ok(());
+        }
+
+        self.set_codebase_snapshot(latest).await;
+        Ok(())
+    }
+
+    pub(crate) async fn refresh_codebase_snapshot(
+        &self,
+        turn_context: &TurnContext,
+    ) -> anyhow::Result<()> {
+        let snapshot = CodebaseSnapshot::capture(turn_context.cwd.clone()).await?;
+        self.set_codebase_snapshot(snapshot).await;
+        Ok(())
+    }
+
     pub(crate) fn build_initial_context(&self, turn_context: &TurnContext) -> Vec<ResponseItem> {
         let mut items = Vec::<ResponseItem>::with_capacity(2);
         if let Some(user_instructions) = turn_context.user_instructions.as_deref() {
@@ -1191,12 +1293,7 @@ async fn submission_loop(
 
                 let tools_config = ToolsConfig::new(&ToolsConfigParams {
                     model_family: &effective_family,
-                    include_plan_tool: config.include_plan_tool,
-                    include_apply_patch_tool: config.include_apply_patch_tool,
-                    include_web_search_request: config.tools_web_search_request,
-                    use_streamable_shell_tool: config.use_experimental_streamable_shell_tool,
-                    include_view_image_tool: config.include_view_image_tool,
-                    experimental_unified_exec_tool: config.use_experimental_unified_exec_tool,
+                    features: &config.features,
                 });
 
                 let new_turn_context = TurnContext {
@@ -1293,14 +1390,7 @@ async fn submission_loop(
                         client,
                         tools_config: ToolsConfig::new(&ToolsConfigParams {
                             model_family: &model_family,
-                            include_plan_tool: config.include_plan_tool,
-                            include_apply_patch_tool: config.include_apply_patch_tool,
-                            include_web_search_request: config.tools_web_search_request,
-                            use_streamable_shell_tool: config
-                                .use_experimental_streamable_shell_tool,
-                            include_view_image_tool: config.include_view_image_tool,
-                            experimental_unified_exec_tool: config
-                                .use_experimental_unified_exec_tool,
+                            features: &config.features,
                         }),
                         user_instructions: turn_context.user_instructions.clone(),
                         base_instructions: turn_context.base_instructions.clone(),
@@ -1400,10 +1490,18 @@ async fn submission_loop(
 
                 // This is a cheap lookup from the connection manager's cache.
                 let tools = sess.services.mcp_connection_manager.list_all_tools();
+                let auth_statuses = compute_auth_statuses(
+                    config.mcp_servers.iter(),
+                    config.mcp_oauth_credentials_store_mode,
+                )
+                .await;
                 let event = Event {
                     id: sub_id,
                     msg: EventMsg::McpListToolsResponse(
-                        crate::protocol::McpListToolsResponseEvent { tools },
+                        crate::protocol::McpListToolsResponseEvent {
+                            tools,
+                            auth_statuses,
+                        },
                     ),
                 };
                 sess.send_event(event).await;
@@ -1524,14 +1622,15 @@ async fn spawn_review_thread(
     let model = config.review_model.clone();
     let review_model_family = find_family_for_model(&model)
         .unwrap_or_else(|| parent_turn_context.client.get_model_family());
+    // For reviews, disable plan, web_search, view_image regardless of global settings.
+    let mut review_features = config.features.clone();
+    review_features.disable(crate::features::Feature::PlanTool);
+    review_features.disable(crate::features::Feature::WebSearchRequest);
+    review_features.disable(crate::features::Feature::ViewImageTool);
+    review_features.disable(crate::features::Feature::StreamableShell);
     let tools_config = ToolsConfig::new(&ToolsConfigParams {
         model_family: &review_model_family,
-        include_plan_tool: false,
-        include_apply_patch_tool: config.include_apply_patch_tool,
-        include_web_search_request: false,
-        use_streamable_shell_tool: false,
-        include_view_image_tool: false,
-        experimental_unified_exec_tool: config.use_experimental_unified_exec_tool,
+        features: &review_features,
     });
 
     let base_instructions = REVIEW_PROMPT.to_string();
@@ -1584,7 +1683,7 @@ async fn spawn_review_thread(
 
     // Seed the child task with the review prompt as the initial user message.
     let input: Vec<InputItem> = vec![InputItem::Text {
-        text: format!("{base_instructions}\n\n---\n\nNow, here's your task: {review_prompt}"),
+        text: review_prompt,
     }];
     let tc = Arc::new(review_turn_context);
 
@@ -1622,6 +1721,7 @@ pub(crate) async fn run_task(
     turn_context: Arc<TurnContext>,
     sub_id: String,
     input: Vec<InputItem>,
+    task_kind: TaskKind,
 ) -> Option<String> {
     if input.is_empty() {
         return None;
@@ -1649,6 +1749,14 @@ pub(crate) async fn run_task(
             .await;
     }
 
+    if !is_review_mode
+        && let Err(err) = sess
+            .emit_codebase_delta_if_changed(turn_context.as_ref(), &sub_id)
+            .await
+    {
+        warn!(error = ?err, "failed to compute codebase changes");
+    }
+
     let mut last_agent_message: Option<String> = None;
     // Although from the perspective of codex.rs, TurnDiffTracker has the lifecycle of a Task which contains
     // many turns, from the perspective of the user, it is a single turn.
@@ -1705,6 +1813,7 @@ pub(crate) async fn run_task(
             Arc::clone(&turn_diff_tracker),
             sub_id.clone(),
             turn_input,
+            task_kind,
         )
         .await
         {
@@ -1857,6 +1966,7 @@ pub(crate) async fn run_task(
                     );
                     sess.notifier()
                         .notify(&UserNotification::AgentTurnComplete {
+                            thread_id: sess.conversation_id.to_string(),
                             turn_id: sub_id.clone(),
                             input_messages: turn_input_messages,
                             last_assistant_message: last_agent_message.clone(),
@@ -1880,6 +1990,11 @@ pub(crate) async fn run_task(
         }
     }
 
+    if !is_review_mode && let Err(err) = sess.refresh_codebase_snapshot(turn_context.as_ref()).await
+    {
+        warn!(error = ?err, "failed to refresh codebase snapshot");
+    }
+
     // If this was a review thread and we have a final assistant message,
     // try to parse it as a ReviewOutput.
     //
@@ -1930,6 +2045,7 @@ async fn run_turn(
     turn_diff_tracker: SharedTurnDiffTracker,
     sub_id: String,
     input: Vec<ResponseItem>,
+    task_kind: TaskKind,
 ) -> CodexResult<TurnRunResult> {
     let mcp_tools = sess.services.mcp_connection_manager.list_all_tools();
     let router = Arc::new(ToolRouter::from_config(
@@ -1959,6 +2075,7 @@ async fn run_turn(
             Arc::clone(&turn_diff_tracker),
             &sub_id,
             &prompt,
+            task_kind,
         )
         .await
         {
@@ -1996,9 +2113,7 @@ async fn run_turn(
                     // at a seemingly frozen screen.
                     sess.notify_stream_error(
                         &sub_id,
-                        format!(
-                            "stream error: {e}; retrying {retries}/{max_retries} in {delay:?}…"
-                        ),
+                        format!("Re-connecting... {retries}/{max_retries}"),
                     )
                     .await;
 
@@ -2034,6 +2149,7 @@ async fn try_run_turn(
     turn_diff_tracker: SharedTurnDiffTracker,
     sub_id: &str,
     prompt: &Prompt,
+    task_kind: TaskKind,
 ) -> CodexResult<TurnRunResult> {
     // call_ids that are part of this response.
     let completed_call_ids = prompt
@@ -2099,16 +2215,21 @@ async fn try_run_turn(
         summary: turn_context.client.get_reasoning_summary(),
     });
     sess.persist_rollout_items(&[rollout_item]).await;
-    let mut stream = turn_context.client.clone().stream(&prompt).await?;
+    let mut stream = turn_context
+        .client
+        .clone()
+        .stream_with_task_kind(prompt.as_ref(), task_kind)
+        .await?;
 
-    let mut output = Vec::new();
-    let mut tool_runtime = ToolCallRuntime::new(
+    let tool_runtime = ToolCallRuntime::new(
         Arc::clone(&router),
         Arc::clone(&sess),
         Arc::clone(&turn_context),
         Arc::clone(&turn_diff_tracker),
         sub_id.to_string(),
     );
+    let mut output: FuturesOrdered<BoxFuture<CodexResult<ProcessedResponseItem>>> =
+        FuturesOrdered::new();
 
     loop {
         // Poll the next item from the model stream. We must inspect *both* Ok and Err
@@ -2116,9 +2237,8 @@ async fn try_run_turn(
         // `response.completed`) bubble up and trigger the caller's retry logic.
         let event = stream.next().await;
         let event = match event {
-            Some(event) => event,
+            Some(res) => res?,
             None => {
-                tool_runtime.abort_all();
                 return Err(CodexErr::Stream(
                     "stream closed before response.completed".into(),
                     None,
@@ -2126,14 +2246,8 @@ async fn try_run_turn(
             }
         };
 
-        let event = match event {
-            Ok(ev) => ev,
-            Err(e) => {
-                tool_runtime.abort_all();
-                // Propagate the underlying stream error to the caller (run_turn), which
-                // will apply the configured `stream_max_retries` policy.
-                return Err(e);
-            }
+        let add_completed = &mut |response_item: ProcessedResponseItem| {
+            output.push_back(future::ready(Ok(response_item)).boxed());
         };
 
         match event {
@@ -2143,14 +2257,18 @@ async fn try_run_turn(
                     Ok(Some(call)) => {
                         let payload_preview = call.payload.log_payload().into_owned();
                         tracing::info!("ToolCall: {} {}", call.tool_name, payload_preview);
-                        let index = output.len();
-                        output.push(ProcessedResponseItem {
-                            item,
-                            response: None,
-                        });
-                        tool_runtime
-                            .handle_tool_call(call, index, output.as_mut_slice())
-                            .await?;
+
+                        let response = tool_runtime.handle_tool_call(call);
+
+                        output.push_back(
+                            async move {
+                                Ok(ProcessedResponseItem {
+                                    item,
+                                    response: Some(response.await?),
+                                })
+                            }
+                            .boxed(),
+                        );
                     }
                     Ok(None) => {
                         let response = handle_non_tool_response_item(
@@ -2160,7 +2278,7 @@ async fn try_run_turn(
                             item.clone(),
                         )
                         .await?;
-                        output.push(ProcessedResponseItem { item, response });
+                        add_completed(ProcessedResponseItem { item, response });
                     }
                     Err(FunctionCallError::MissingLocalShellCallId) => {
                         let msg = "LocalShellCall without call_id or id";
@@ -2177,7 +2295,7 @@ async fn try_run_turn(
                                 success: None,
                             },
                         };
-                        output.push(ProcessedResponseItem {
+                        add_completed(ProcessedResponseItem {
                             item,
                             response: Some(response),
                         });
@@ -2190,7 +2308,7 @@ async fn try_run_turn(
                                 success: None,
                             },
                         };
-                        output.push(ProcessedResponseItem {
+                        add_completed(ProcessedResponseItem {
                             item,
                             response: Some(response),
                         });
@@ -2221,7 +2339,7 @@ async fn try_run_turn(
                 sess.update_token_usage_info(sub_id, turn_context.as_ref(), token_usage.as_ref())
                     .await;
 
-                tool_runtime.resolve_pending(output.as_mut_slice()).await?;
+                let processed_items: Vec<ProcessedResponseItem> = output.try_collect().await?;
 
                 let unified_diff = {
                     let mut tracker = turn_diff_tracker.lock().await;
@@ -2237,7 +2355,7 @@ async fn try_run_turn(
                 }
 
                 let result = TurnRunResult {
-                    processed_items: output,
+                    processed_items,
                     total_token_usage: token_usage.clone(),
                 };
 
@@ -2740,12 +2858,7 @@ mod tests {
         );
         let tools_config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &config.model_family,
-            include_plan_tool: config.include_plan_tool,
-            include_apply_patch_tool: config.include_apply_patch_tool,
-            include_web_search_request: config.tools_web_search_request,
-            use_streamable_shell_tool: config.use_experimental_streamable_shell_tool,
-            include_view_image_tool: config.include_view_image_tool,
-            experimental_unified_exec_tool: config.use_experimental_unified_exec_tool,
+            features: &config.features,
         });
         let turn_context = TurnContext {
             client,
@@ -2813,12 +2926,7 @@ mod tests {
         );
         let tools_config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &config.model_family,
-            include_plan_tool: config.include_plan_tool,
-            include_apply_patch_tool: config.include_apply_patch_tool,
-            include_web_search_request: config.tools_web_search_request,
-            use_streamable_shell_tool: config.use_experimental_streamable_shell_tool,
-            include_view_image_tool: config.include_view_image_tool,
-            experimental_unified_exec_tool: config.use_experimental_unified_exec_tool,
+            features: &config.features,
         });
         let turn_context = Arc::new(TurnContext {
             client,

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;
@@ -93,17 +90,36 @@ 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 {
@@ -121,9 +137,7 @@ async fn run_compact_task_inner(
                     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;
@@ -245,7 +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/config.rs

@@ -17,6 +17,10 @@ use crate::config_types::ShellEnvironmentPolicy;
 use crate::config_types::ShellEnvironmentPolicyToml;
 use crate::config_types::Tui;
 use crate::config_types::UriBasedFileOpener;
+use crate::features::Feature;
+use crate::features::FeatureOverrides;
+use crate::features::Features;
+use crate::features::FeaturesToml;
 use crate::git_info::resolve_root_git_project_for_trust;
 use crate::model_family::ModelFamily;
 use crate::model_family::derive_default_model_family;
@@ -33,12 +37,15 @@ 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_rmcp_client::OAuthCredentialsStoreMode;
 use dirs::home_dir;
 use serde::Deserialize;
 use std::collections::BTreeMap;
 use std::collections::HashMap;
+use std::io::ErrorKind;
 use std::path::Path;
 use std::path::PathBuf;
+
 use tempfile::NamedTempFile;
 use toml::Value as TomlValue;
 use toml_edit::Array as TomlArray;
@@ -142,6 +149,15 @@ pub struct Config {
     /// Definition for MCP servers that Codex can reach out to for tool calls.
     pub mcp_servers: HashMap<String, McpServerConfig>,
 
+    /// Preferred store for MCP OAuth credentials.
+    /// keyring: Use an OS-specific keyring service.
+    ///          Credentials stored in the keyring will only be readable by Codex unless the user explicitly grants access via OS-level keyring access.
+    ///          https://github.com/openai/codex/blob/main/codex-rs/rmcp-client/src/oauth.rs#L2
+    /// file: CODEX_HOME/.credentials.json
+    ///       This file will be readable to Codex and other applications running as the same user.
+    /// auto (default): keyring if available, otherwise file.
+    pub mcp_oauth_credentials_store_mode: OAuthCredentialsStoreMode,
+
     /// Combined provider map (defaults merged with user-defined overrides).
     pub model_providers: HashMap<String, ModelProviderInfo>,
 
@@ -206,6 +222,9 @@ pub struct Config {
     /// Include the `view_image` tool that lets the agent attach a local image path to context.
     pub include_view_image_tool: bool,
 
+    /// Centralized feature flags; source of truth for feature gating.
+    pub features: Features,
+
     /// The active profile name used to derive this `Config` (if any).
     pub active_profile: Option<String>,
 
@@ -301,12 +320,35 @@ pub async fn load_global_mcp_servers(
         return Ok(BTreeMap::new());
     };
 
+    ensure_no_inline_bearer_tokens(servers_value)?;
+
     servers_value
         .clone()
         .try_into()
         .map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidData, e))
 }
 
+/// We briefly allowed plain text bearer_token fields in MCP server configs.
+/// We want to warn people who recently added these fields but can remove this after a few months.
+fn ensure_no_inline_bearer_tokens(value: &TomlValue) -> std::io::Result<()> {
+    let Some(servers_table) = value.as_table() else {
+        return Ok(());
+    };
+
+    for (server_name, server_value) in servers_table {
+        if let Some(server_table) = server_value.as_table()
+            && server_table.contains_key("bearer_token")
+        {
+            let message = format!(
+                "mcp_servers.{server_name} uses unsupported `bearer_token`; set `bearer_token_env_var`."
+            );
+            return Err(std::io::Error::new(ErrorKind::InvalidData, message));
+        }
+    }
+
+    Ok(())
+}
+
 pub fn write_global_mcp_servers(
     codex_home: &Path,
     servers: &BTreeMap<String, McpServerConfig>,
@@ -355,14 +397,21 @@ pub fn write_global_mcp_servers(
                         entry["env"] = TomlItem::Table(env_table);
                     }
                 }
-                McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
+                McpServerTransportConfig::StreamableHttp {
+                    url,
+                    bearer_token_env_var,
+                } => {
                     entry["url"] = toml_edit::value(url.clone());
-                    if let Some(token) = bearer_token {
-                        entry["bearer_token"] = toml_edit::value(token.clone());
+                    if let Some(env_var) = bearer_token_env_var {
+                        entry["bearer_token_env_var"] = toml_edit::value(env_var.clone());
                     }
                 }
             }
 
+            if !config.enabled {
+                entry["enabled"] = toml_edit::value(false);
+            }
+
             if let Some(timeout) = config.startup_timeout_sec {
                 entry["startup_timeout_sec"] = toml_edit::value(timeout.as_secs_f64());
             }
@@ -694,6 +743,14 @@ pub struct ConfigToml {
     #[serde(default)]
     pub mcp_servers: HashMap<String, McpServerConfig>,
 
+    /// Preferred backend for storing MCP OAuth credentials.
+    /// keyring: Use an OS-specific keyring service.
+    ///          https://github.com/openai/codex/blob/main/codex-rs/rmcp-client/src/oauth.rs#L2
+    /// file: Use a file in the Codex home directory.
+    /// auto (default): Use the OS-specific keyring service if available, otherwise use a file.
+    #[serde(default)]
+    pub mcp_oauth_credentials_store: Option<OAuthCredentialsStoreMode>,
+
     /// User-defined provider entries that extend/override the built-in list.
     #[serde(default)]
     pub model_providers: HashMap<String, ModelProviderInfo>,
@@ -744,19 +801,15 @@ pub struct ConfigToml {
     /// Base URL for requests to ChatGPT (as opposed to the OpenAI API).
     pub chatgpt_base_url: Option<String>,
 
-    /// Experimental path to a file whose contents replace the built-in BASE_INSTRUCTIONS.
-    pub experimental_instructions_file: Option<PathBuf>,
-
-    pub experimental_use_exec_command_tool: Option<bool>,
-    pub experimental_use_unified_exec_tool: Option<bool>,
-    pub experimental_use_rmcp_client: Option<bool>,
-    pub experimental_use_freeform_apply_patch: Option<bool>,
-
     pub projects: Option<HashMap<String, ProjectConfig>>,
 
     /// Nested tools section for feature toggles
     pub tools: Option<ToolsToml>,
 
+    /// Centralized feature flags (new). Prefer this over individual toggles.
+    #[serde(default)]
+    pub features: Option<FeaturesToml>,
+
     /// When true, disables burst-paste detection for typed input entirely.
     /// All characters are inserted as they are received, and no buffering
     /// or placeholder replacement will occur for fast keypress bursts.
@@ -767,6 +820,13 @@ pub struct ConfigToml {
 
     /// Tracks whether the Windows onboarding screen has been acknowledged.
     pub windows_wsl_setup_acknowledged: Option<bool>,
+
+    /// Legacy, now use features
+    pub experimental_instructions_file: Option<PathBuf>,
+    pub experimental_use_exec_command_tool: Option<bool>,
+    pub experimental_use_unified_exec_tool: Option<bool>,
+    pub experimental_use_rmcp_client: Option<bool>,
+    pub experimental_use_freeform_apply_patch: Option<bool>,
 }
 
 impl From<ConfigToml> for UserSavedConfig {
@@ -930,9 +990,9 @@ impl Config {
             config_profile: config_profile_key,
             codex_linux_sandbox_exe,
             base_instructions,
-            include_plan_tool,
-            include_apply_patch_tool,
-            include_view_image_tool,
+            include_plan_tool: include_plan_tool_override,
+            include_apply_patch_tool: include_apply_patch_tool_override,
+            include_view_image_tool: include_view_image_tool_override,
             show_raw_agent_reasoning,
             tools_web_search_request: override_tools_web_search_request,
         } = overrides;
@@ -955,6 +1015,15 @@ impl Config {
             None => ConfigProfile::default(),
         };
 
+        let feature_overrides = FeatureOverrides {
+            include_plan_tool: include_plan_tool_override,
+            include_apply_patch_tool: include_apply_patch_tool_override,
+            include_view_image_tool: include_view_image_tool_override,
+            web_search_request: override_tools_web_search_request,
+        };
+
+        let features = Features::from_config(&cfg, &config_profile, feature_overrides);
+
         let sandbox_policy = cfg.derive_sandbox_policy(sandbox_mode);
 
         let mut model_providers = built_in_model_providers();
@@ -1000,13 +1069,13 @@ impl Config {
 
         let history = cfg.history.unwrap_or_default();
 
-        let tools_web_search_request = override_tools_web_search_request
-            .or(cfg.tools.as_ref().and_then(|t| t.web_search))
-            .unwrap_or(false);
-
-        let include_view_image_tool = include_view_image_tool
-            .or(cfg.tools.as_ref().and_then(|t| t.view_image))
-            .unwrap_or(true);
+        let include_plan_tool_flag = features.enabled(Feature::PlanTool);
+        let include_apply_patch_tool_flag = features.enabled(Feature::ApplyPatchFreeform);
+        let include_view_image_tool_flag = features.enabled(Feature::ViewImageTool);
+        let tools_web_search_request = features.enabled(Feature::WebSearchRequest);
+        let use_experimental_streamable_shell_tool = features.enabled(Feature::StreamableShell);
+        let use_experimental_unified_exec_tool = features.enabled(Feature::UnifiedExec);
+        let use_experimental_use_rmcp_client = features.enabled(Feature::RmcpClient);
 
         let model = model
             .or(config_profile.model)
@@ -1074,6 +1143,9 @@ impl Config {
             user_instructions,
             base_instructions,
             mcp_servers: cfg.mcp_servers,
+            // The config.toml omits "_mode" because it's a config file. However, "_mode"
+            // is important in code to differentiate the mode from the store implementation.
+            mcp_oauth_credentials_store_mode: cfg.mcp_oauth_credentials_store.unwrap_or_default(),
             model_providers,
             project_doc_max_bytes: cfg.project_doc_max_bytes.unwrap_or(PROJECT_DOC_MAX_BYTES),
             project_doc_fallback_filenames: cfg
@@ -1111,19 +1183,14 @@ impl Config {
                 .chatgpt_base_url
                 .or(cfg.chatgpt_base_url)
                 .unwrap_or("https://chatgpt.com/backend-api/".to_string()),
-            include_plan_tool: include_plan_tool.unwrap_or(false),
-            include_apply_patch_tool: include_apply_patch_tool
-                .or(cfg.experimental_use_freeform_apply_patch)
-                .unwrap_or(false),
+            include_plan_tool: include_plan_tool_flag,
+            include_apply_patch_tool: include_apply_patch_tool_flag,
             tools_web_search_request,
-            use_experimental_streamable_shell_tool: cfg
-                .experimental_use_exec_command_tool
-                .unwrap_or(false),
-            use_experimental_unified_exec_tool: cfg
-                .experimental_use_unified_exec_tool
-                .unwrap_or(false),
-            use_experimental_use_rmcp_client: cfg.experimental_use_rmcp_client.unwrap_or(false),
-            include_view_image_tool,
+            use_experimental_streamable_shell_tool,
+            use_experimental_unified_exec_tool,
+            use_experimental_use_rmcp_client,
+            include_view_image_tool: include_view_image_tool_flag,
+            features,
             active_profile: active_profile_name,
             windows_wsl_setup_acknowledged: cfg.windows_wsl_setup_acknowledged.unwrap_or(false),
             disable_paste_burst: cfg.disable_paste_burst.unwrap_or(false),
@@ -1256,6 +1323,7 @@ pub fn log_dir(cfg: &Config) -> std::io::Result<PathBuf> {
 mod tests {
     use crate::config_types::HistoryPersistence;
     use crate::config_types::Notifications;
+    use crate::features::Feature;
 
     use super::*;
     use pretty_assertions::assert_eq;
@@ -1364,6 +1432,172 @@ exclude_slash_tmp = true
         );
     }
 
+    #[test]
+    fn config_defaults_to_auto_oauth_store_mode() -> std::io::Result<()> {
+        let codex_home = TempDir::new()?;
+        let cfg = ConfigToml::default();
+
+        let config = Config::load_from_base_config_with_overrides(
+            cfg,
+            ConfigOverrides::default(),
+            codex_home.path().to_path_buf(),
+        )?;
+
+        assert_eq!(
+            config.mcp_oauth_credentials_store_mode,
+            OAuthCredentialsStoreMode::Auto,
+        );
+
+        Ok(())
+    }
+
+    #[test]
+    fn profile_legacy_toggles_override_base() -> std::io::Result<()> {
+        let codex_home = TempDir::new()?;
+        let mut profiles = HashMap::new();
+        profiles.insert(
+            "work".to_string(),
+            ConfigProfile {
+                include_plan_tool: Some(true),
+                include_view_image_tool: Some(false),
+                ..Default::default()
+            },
+        );
+        let cfg = ConfigToml {
+            profiles,
+            profile: Some("work".to_string()),
+            ..Default::default()
+        };
+
+        let config = Config::load_from_base_config_with_overrides(
+            cfg,
+            ConfigOverrides::default(),
+            codex_home.path().to_path_buf(),
+        )?;
+
+        assert!(config.features.enabled(Feature::PlanTool));
+        assert!(!config.features.enabled(Feature::ViewImageTool));
+        assert!(config.include_plan_tool);
+        assert!(!config.include_view_image_tool);
+
+        Ok(())
+    }
+
+    #[test]
+    fn feature_table_overrides_legacy_flags() -> std::io::Result<()> {
+        let codex_home = TempDir::new()?;
+        let mut entries = BTreeMap::new();
+        entries.insert("plan_tool".to_string(), false);
+        entries.insert("apply_patch_freeform".to_string(), false);
+        let cfg = ConfigToml {
+            features: Some(crate::features::FeaturesToml { entries }),
+            ..Default::default()
+        };
+
+        let config = Config::load_from_base_config_with_overrides(
+            cfg,
+            ConfigOverrides::default(),
+            codex_home.path().to_path_buf(),
+        )?;
+
+        assert!(!config.features.enabled(Feature::PlanTool));
+        assert!(!config.features.enabled(Feature::ApplyPatchFreeform));
+        assert!(!config.include_plan_tool);
+        assert!(!config.include_apply_patch_tool);
+
+        Ok(())
+    }
+
+    #[test]
+    fn legacy_toggles_map_to_features() -> std::io::Result<()> {
+        let codex_home = TempDir::new()?;
+        let cfg = ConfigToml {
+            experimental_use_exec_command_tool: Some(true),
+            experimental_use_unified_exec_tool: Some(true),
+            experimental_use_rmcp_client: Some(true),
+            experimental_use_freeform_apply_patch: Some(true),
+            ..Default::default()
+        };
+
+        let config = Config::load_from_base_config_with_overrides(
+            cfg,
+            ConfigOverrides::default(),
+            codex_home.path().to_path_buf(),
+        )?;
+
+        assert!(config.features.enabled(Feature::ApplyPatchFreeform));
+        assert!(config.features.enabled(Feature::StreamableShell));
+        assert!(config.features.enabled(Feature::UnifiedExec));
+        assert!(config.features.enabled(Feature::RmcpClient));
+
+        assert!(config.include_apply_patch_tool);
+        assert!(config.use_experimental_streamable_shell_tool);
+        assert!(config.use_experimental_unified_exec_tool);
+        assert!(config.use_experimental_use_rmcp_client);
+
+        Ok(())
+    }
+
+    #[test]
+    fn config_honors_explicit_file_oauth_store_mode() -> std::io::Result<()> {
+        let codex_home = TempDir::new()?;
+        let cfg = ConfigToml {
+            mcp_oauth_credentials_store: Some(OAuthCredentialsStoreMode::File),
+            ..Default::default()
+        };
+
+        let config = Config::load_from_base_config_with_overrides(
+            cfg,
+            ConfigOverrides::default(),
+            codex_home.path().to_path_buf(),
+        )?;
+
+        assert_eq!(
+            config.mcp_oauth_credentials_store_mode,
+            OAuthCredentialsStoreMode::File,
+        );
+
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn managed_config_overrides_oauth_store_mode() -> anyhow::Result<()> {
+        let codex_home = TempDir::new()?;
+        let managed_path = codex_home.path().join("managed_config.toml");
+        let config_path = codex_home.path().join(CONFIG_TOML_FILE);
+
+        std::fs::write(&config_path, "mcp_oauth_credentials_store = \"file\"\n")?;
+        std::fs::write(&managed_path, "mcp_oauth_credentials_store = \"keyring\"\n")?;
+
+        let overrides = crate::config_loader::LoaderOverrides {
+            managed_config_path: Some(managed_path.clone()),
+            #[cfg(target_os = "macos")]
+            managed_preferences_base64: None,
+        };
+
+        let root_value = load_resolved_config(codex_home.path(), Vec::new(), overrides).await?;
+        let cfg: ConfigToml = root_value.try_into().map_err(|e| {
+            tracing::error!("Failed to deserialize overridden config: {e}");
+            std::io::Error::new(std::io::ErrorKind::InvalidData, e)
+        })?;
+        assert_eq!(
+            cfg.mcp_oauth_credentials_store,
+            Some(OAuthCredentialsStoreMode::Keyring),
+        );
+
+        let final_config = Config::load_from_base_config_with_overrides(
+            cfg,
+            ConfigOverrides::default(),
+            codex_home.path().to_path_buf(),
+        )?;
+        assert_eq!(
+            final_config.mcp_oauth_credentials_store_mode,
+            OAuthCredentialsStoreMode::Keyring,
+        );
+
+        Ok(())
+    }
+
     #[tokio::test]
     async fn load_global_mcp_servers_returns_empty_if_missing() -> anyhow::Result<()> {
         let codex_home = TempDir::new()?;
@@ -1387,6 +1621,7 @@ exclude_slash_tmp = true
                     args: vec!["hello".to_string()],
                     env: None,
                 },
+                enabled: true,
                 startup_timeout_sec: Some(Duration::from_secs(3)),
                 tool_timeout_sec: Some(Duration::from_secs(5)),
             },
@@ -1407,6 +1642,7 @@ exclude_slash_tmp = true
         }
         assert_eq!(docs.startup_timeout_sec, Some(Duration::from_secs(3)));
         assert_eq!(docs.tool_timeout_sec, Some(Duration::from_secs(5)));
+        assert!(docs.enabled);
 
         let empty = BTreeMap::new();
         write_global_mcp_servers(codex_home.path(), &empty)?;
@@ -1471,6 +1707,31 @@ startup_timeout_ms = 2500
         Ok(())
     }
 
+    #[tokio::test]
+    async fn load_global_mcp_servers_rejects_inline_bearer_token() -> anyhow::Result<()> {
+        let codex_home = TempDir::new()?;
+        let config_path = codex_home.path().join(CONFIG_TOML_FILE);
+
+        std::fs::write(
+            &config_path,
+            r#"
+[mcp_servers.docs]
+url = "https://example.com/mcp"
+bearer_token = "secret"
+"#,
+        )?;
+
+        let err = load_global_mcp_servers(codex_home.path())
+            .await
+            .expect_err("bearer_token entries should be rejected");
+
+        assert_eq!(err.kind(), std::io::ErrorKind::InvalidData);
+        assert!(err.to_string().contains("bearer_token"));
+        assert!(err.to_string().contains("bearer_token_env_var"));
+
+        Ok(())
+    }
+
     #[tokio::test]
     async fn write_global_mcp_servers_serializes_env_sorted() -> anyhow::Result<()> {
         let codex_home = TempDir::new()?;
@@ -1486,6 +1747,7 @@ startup_timeout_ms = 2500
                         ("ALPHA_VAR".to_string(), "1".to_string()),
                     ])),
                 },
+                enabled: true,
                 startup_timeout_sec: None,
                 tool_timeout_sec: None,
             },
@@ -1534,8 +1796,9 @@ ZIG_VAR = "3"
             McpServerConfig {
                 transport: McpServerTransportConfig::StreamableHttp {
                     url: "https://example.com/mcp".to_string(),
-                    bearer_token: Some("secret-token".to_string()),
+                    bearer_token_env_var: Some("MCP_TOKEN".to_string()),
                 },
+                enabled: true,
                 startup_timeout_sec: Some(Duration::from_secs(2)),
                 tool_timeout_sec: None,
             },
@@ -1549,7 +1812,7 @@ ZIG_VAR = "3"
             serialized,
             r#"[mcp_servers.docs]
 url = "https://example.com/mcp"
-bearer_token = "secret-token"
+bearer_token_env_var = "MCP_TOKEN"
 startup_timeout_sec = 2.0
 "#
         );
@@ -1557,9 +1820,12 @@ startup_timeout_sec = 2.0
         let loaded = load_global_mcp_servers(codex_home.path()).await?;
         let docs = loaded.get("docs").expect("docs entry");
         match &docs.transport {
-            McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
+            McpServerTransportConfig::StreamableHttp {
+                url,
+                bearer_token_env_var,
+            } => {
                 assert_eq!(url, "https://example.com/mcp");
-                assert_eq!(bearer_token.as_deref(), Some("secret-token"));
+                assert_eq!(bearer_token_env_var.as_deref(), Some("MCP_TOKEN"));
             }
             other => panic!("unexpected transport {other:?}"),
         }
@@ -1570,8 +1836,9 @@ startup_timeout_sec = 2.0
             McpServerConfig {
                 transport: McpServerTransportConfig::StreamableHttp {
                     url: "https://example.com/mcp".to_string(),
-                    bearer_token: None,
+                    bearer_token_env_var: None,
                 },
+                enabled: true,
                 startup_timeout_sec: None,
                 tool_timeout_sec: None,
             },
@@ -1589,9 +1856,12 @@ url = "https://example.com/mcp"
         let loaded = load_global_mcp_servers(codex_home.path()).await?;
         let docs = loaded.get("docs").expect("docs entry");
         match &docs.transport {
-            McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
+            McpServerTransportConfig::StreamableHttp {
+                url,
+                bearer_token_env_var,
+            } => {
                 assert_eq!(url, "https://example.com/mcp");
-                assert!(bearer_token.is_none());
+                assert!(bearer_token_env_var.is_none());
             }
             other => panic!("unexpected transport {other:?}"),
         }
@@ -1599,6 +1869,40 @@ url = "https://example.com/mcp"
         Ok(())
     }
 
+    #[tokio::test]
+    async fn write_global_mcp_servers_serializes_disabled_flag() -> anyhow::Result<()> {
+        let codex_home = TempDir::new()?;
+
+        let servers = BTreeMap::from([(
+            "docs".to_string(),
+            McpServerConfig {
+                transport: McpServerTransportConfig::Stdio {
+                    command: "docs-server".to_string(),
+                    args: Vec::new(),
+                    env: None,
+                },
+                enabled: false,
+                startup_timeout_sec: None,
+                tool_timeout_sec: None,
+            },
+        )]);
+
+        write_global_mcp_servers(codex_home.path(), &servers)?;
+
+        let config_path = codex_home.path().join(CONFIG_TOML_FILE);
+        let serialized = std::fs::read_to_string(&config_path)?;
+        assert!(
+            serialized.contains("enabled = false"),
+            "serialized config missing disabled flag:\n{serialized}"
+        );
+
+        let loaded = load_global_mcp_servers(codex_home.path()).await?;
+        let docs = loaded.get("docs").expect("docs entry");
+        assert!(!docs.enabled);
+
+        Ok(())
+    }
+
     #[tokio::test]
     async fn persist_model_selection_updates_defaults() -> anyhow::Result<()> {
         let codex_home = TempDir::new()?;
@@ -1896,6 +2200,7 @@ model_verbosity = "high"
                 notify: None,
                 cwd: fixture.cwd(),
                 mcp_servers: HashMap::new(),
+                mcp_oauth_credentials_store_mode: Default::default(),
                 model_providers: fixture.model_provider_map.clone(),
                 project_doc_max_bytes: PROJECT_DOC_MAX_BYTES,
                 project_doc_fallback_filenames: Vec::new(),
@@ -1917,6 +2222,7 @@ model_verbosity = "high"
                 use_experimental_unified_exec_tool: false,
                 use_experimental_use_rmcp_client: false,
                 include_view_image_tool: true,
+                features: Features::with_defaults(),
                 active_profile: Some("o3".to_string()),
                 windows_wsl_setup_acknowledged: false,
                 disable_paste_burst: false,
@@ -1958,6 +2264,7 @@ model_verbosity = "high"
             notify: None,
             cwd: fixture.cwd(),
             mcp_servers: HashMap::new(),
+            mcp_oauth_credentials_store_mode: Default::default(),
             model_providers: fixture.model_provider_map.clone(),
             project_doc_max_bytes: PROJECT_DOC_MAX_BYTES,
             project_doc_fallback_filenames: Vec::new(),
@@ -1979,6 +2286,7 @@ model_verbosity = "high"
             use_experimental_unified_exec_tool: false,
             use_experimental_use_rmcp_client: false,
             include_view_image_tool: true,
+            features: Features::with_defaults(),
             active_profile: Some("gpt3".to_string()),
             windows_wsl_setup_acknowledged: false,
             disable_paste_burst: false,
@@ -2035,6 +2343,7 @@ model_verbosity = "high"
             notify: None,
             cwd: fixture.cwd(),
             mcp_servers: HashMap::new(),
+            mcp_oauth_credentials_store_mode: Default::default(),
             model_providers: fixture.model_provider_map.clone(),
             project_doc_max_bytes: PROJECT_DOC_MAX_BYTES,
             project_doc_fallback_filenames: Vec::new(),
@@ -2056,6 +2365,7 @@ model_verbosity = "high"
             use_experimental_unified_exec_tool: false,
             use_experimental_use_rmcp_client: false,
             include_view_image_tool: true,
+            features: Features::with_defaults(),
             active_profile: Some("zdr".to_string()),
             windows_wsl_setup_acknowledged: false,
             disable_paste_burst: false,
@@ -2098,6 +2408,7 @@ model_verbosity = "high"
             notify: None,
             cwd: fixture.cwd(),
             mcp_servers: HashMap::new(),
+            mcp_oauth_credentials_store_mode: Default::default(),
             model_providers: fixture.model_provider_map.clone(),
             project_doc_max_bytes: PROJECT_DOC_MAX_BYTES,
             project_doc_fallback_filenames: Vec::new(),
@@ -2119,6 +2430,7 @@ model_verbosity = "high"
             use_experimental_unified_exec_tool: false,
             use_experimental_use_rmcp_client: false,
             include_view_image_tool: true,
+            features: Features::with_defaults(),
             active_profile: Some("gpt5".to_string()),
             windows_wsl_setup_acknowledged: false,
             disable_paste_burst: false,
@@ -2231,6 +2543,7 @@ trust_level = "trusted"
 #[cfg(test)]
 mod notifications_tests {
     use crate::config_types::Notifications;
+    use assert_matches::assert_matches;
     use serde::Deserialize;
 
     #[derive(Deserialize, Debug, PartialEq)]
@@ -2250,10 +2563,7 @@ mod notifications_tests {
             notifications = true
         "#;
         let parsed: RootTomlTest = toml::from_str(toml).expect("deserialize notifications=true");
-        assert!(matches!(
-            parsed.tui.notifications,
-            Notifications::Enabled(true)
-        ));
+        assert_matches!(parsed.tui.notifications, Notifications::Enabled(true));
     }
 
     #[test]
@@ -2264,9 +2574,9 @@ mod notifications_tests {
         "#;
         let parsed: RootTomlTest =
             toml::from_str(toml).expect("deserialize notifications=[\"foo\"]");
-        assert!(matches!(
+        assert_matches!(
             parsed.tui.notifications,
             Notifications::Custom(ref v) if v == &vec!["foo".to_string()]
-        ));
+        );
     }
 }

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,
@@ -48,6 +52,7 @@ impl<'de> Deserialize<'de> for McpServerConfig {
 
             url: Option<String>,
             bearer_token: Option<String>,
+            bearer_token_env_var: Option<String>,
 
             #[serde(default)]
             startup_timeout_sec: Option<f64>,
@@ -55,6 +60,8 @@ impl<'de> Deserialize<'de> for McpServerConfig {
             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)?;
@@ -86,11 +93,15 @@ impl<'de> Deserialize<'de> for McpServerConfig {
                 args,
                 env,
                 url,
-                bearer_token,
+                bearer_token_env_var,
                 ..
             } => {
                 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(),
+                )?;
                 McpServerTransportConfig::Stdio {
                     command,
                     args: args.unwrap_or_default(),
@@ -100,6 +111,7 @@ impl<'de> Deserialize<'de> for McpServerConfig {
             RawMcpServerConfig {
                 url: Some(url),
                 bearer_token,
+                bearer_token_env_var,
                 command,
                 args,
                 env,
@@ -108,7 +120,11 @@ impl<'de> Deserialize<'de> for McpServerConfig {
                 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", "bearer_token", bearer_token.as_ref())?;
+                McpServerTransportConfig::StreamableHttp {
+                    url,
+                    bearer_token_env_var,
+                }
             }
             _ => return Err(SerdeError::custom("invalid transport")),
         };
@@ -117,10 +133,15 @@ 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 {
@@ -135,11 +156,11 @@ pub enum McpServerTransportConfig {
     /// 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>,
     },
 }
 
@@ -450,6 +471,7 @@ mod tests {
                 env: None
             }
         );
+        assert!(cfg.enabled);
     }
 
     #[test]
@@ -470,6 +492,7 @@ mod tests {
                 env: None
             }
         );
+        assert!(cfg.enabled);
     }
 
     #[test]
@@ -491,6 +514,20 @@ mod tests {
                 env: Some(HashMap::from([("FOO".to_string(), "BAR".to_string())]))
             }
         );
+        assert!(cfg.enabled);
+    }
+
+    #[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 +543,18 @@ mod tests {
             cfg.transport,
             McpServerTransportConfig::StreamableHttp {
                 url: "https://example.com/mcp".to_string(),
-                bearer_token: None
+                bearer_token_env_var: 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");
@@ -525,9 +563,10 @@ 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())
             }
         );
+        assert!(cfg.enabled);
     }
 
     #[test]
@@ -553,13 +592,18 @@ mod tests {
     }
 
     #[test]
-    fn deserialize_rejects_bearer_token_for_stdio_transport() {
-        toml::from_str::<McpServerConfig>(
+    fn deserialize_rejects_inline_bearer_token_field() {
+        let err = toml::from_str::<McpServerConfig>(
             r#"
-            command = "echo"
+            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

@@ -210,6 +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;
@@ -283,7 +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/error.rs

@@ -1,6 +1,7 @@
 use crate::exec::ExecToolCallOutput;
 use crate::token_data::KnownPlan;
 use crate::token_data::PlanType;
+use crate::truncate::truncate_middle;
 use codex_protocol::ConversationId;
 use codex_protocol::protocol::RateLimitSnapshot;
 use reqwest::StatusCode;
@@ -12,6 +13,9 @@ 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
@@ -304,21 +308,44 @@ 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 codex_protocol::protocol::RateLimitWindow;
+    use pretty_assertions::assert_eq;
 
     fn rate_limit_snapshot() -> RateLimitSnapshot {
         RateLimitSnapshot {
@@ -348,6 +375,73 @@ 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 {

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

@@ -177,7 +177,7 @@ 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),
                 }));
@@ -195,21 +195,57 @@ pub async fn process_exec_tool_call(
 /// 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.
+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 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
+    const SANDBOX_DENIED_KEYWORDS: [&str; 6] = [
+        "operation not permitted",
+        "permission denied",
+        "read-only file system",
+        "seccomp",
+        "sandbox",
+        "landlock",
+    ];
+
+    if [
+        &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))
+    }) {
+        return 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)]
@@ -436,3 +472,77 @@ 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/executor/backends.rs

@@ -6,6 +6,7 @@ use async_trait::async_trait;
 use crate::CODEX_APPLY_PATCH_ARG1;
 use crate::apply_patch::ApplyPatchExec;
 use crate::exec::ExecParams;
+use crate::executor::ExecutorConfig;
 use crate::function_tool::FunctionCallError;
 
 pub(crate) enum ExecutionMode {
@@ -22,6 +23,7 @@ pub(crate) trait ExecutionBackend: Send + Sync {
         params: ExecParams,
         // Required for downcasting the apply_patch.
         mode: &ExecutionMode,
+        config: &ExecutorConfig,
     ) -> Result<ExecParams, FunctionCallError>;
 
     fn stream_stdout(&self, _mode: &ExecutionMode) -> bool {
@@ -47,6 +49,7 @@ impl ExecutionBackend for ShellBackend {
         &self,
         params: ExecParams,
         mode: &ExecutionMode,
+        _config: &ExecutorConfig,
     ) -> Result<ExecParams, FunctionCallError> {
         match mode {
             ExecutionMode::Shell => Ok(params),
@@ -65,17 +68,22 @@ impl ExecutionBackend for ApplyPatchBackend {
         &self,
         params: ExecParams,
         mode: &ExecutionMode,
+        config: &ExecutorConfig,
     ) -> 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 path_to_codex = if let Some(exe_path) = &config.codex_exe {
+                    exe_path.to_string_lossy().to_string()
+                } else {
+                    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 {

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

@@ -30,19 +30,19 @@ use codex_otel::otel_event_manager::ToolDecisionSource;
 pub(crate) struct ExecutorConfig {
     pub(crate) sandbox_policy: SandboxPolicy,
     pub(crate) sandbox_cwd: PathBuf,
-    codex_linux_sandbox_exe: Option<PathBuf>,
+    pub(crate) codex_exe: Option<PathBuf>,
 }
 
 impl ExecutorConfig {
     pub(crate) fn new(
         sandbox_policy: SandboxPolicy,
         sandbox_cwd: PathBuf,
-        codex_linux_sandbox_exe: Option<PathBuf>,
+        codex_exe: Option<PathBuf>,
     ) -> Self {
         Self {
             sandbox_policy,
             sandbox_cwd,
-            codex_linux_sandbox_exe,
+            codex_exe,
         }
     }
 }
@@ -86,7 +86,14 @@ impl Executor {
                 maybe_translate_shell_command(request.params, session, request.use_shell_profile);
         }
 
-        // Step 1: Normalise parameters via the selected backend.
+        // Step 1: Snapshot sandbox configuration so it stays stable for this run.
+        let config = self
+            .config
+            .read()
+            .map_err(|_| ExecError::rejection("executor config poisoned"))?
+            .clone();
+
+        // Step 2: 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()
@@ -94,16 +101,9 @@ impl Executor {
             None
         };
         request.params = backend
-            .prepare(request.params, &request.mode)
+            .prepare(request.params, &request.mode, &config)
             .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,
@@ -227,7 +227,7 @@ impl Executor {
             sandbox,
             &config.sandbox_policy,
             &config.sandbox_cwd,
-            &config.codex_linux_sandbox_exe,
+            &config.codex_exe,
             stdout_stream,
         )
         .await
@@ -380,6 +380,23 @@ mod tests {
         assert_eq!(message, "failed in sandbox: sandbox stderr");
     }
 
+    #[test]
+    fn sandbox_failure_message_falls_back_to_aggregated_output() {
+        let output = ExecToolCallOutput {
+            exit_code: 101,
+            stdout: StreamOutput::new(String::new()),
+            stderr: StreamOutput::new(String::new()),
+            aggregated_output: StreamOutput::new("aggregate text".to_string()),
+            duration: Duration::from_millis(10),
+            timed_out: false,
+        };
+        let err = SandboxErr::Denied {
+            output: Box::new(output),
+        };
+        let message = sandbox_failure_message(err);
+        assert_eq!(message, "failed in sandbox: aggregate text");
+    }
+
     #[test]
     fn normalize_function_error_synthesizes_payload() {
         let err = FunctionCallError::RespondToModel("boom".to_string());

codex-rs/core/src/features.rs

@@ -0,0 +1,250 @@
+//! 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,
+}
+
+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,
+    },
+];

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/lib.rs

@@ -11,6 +11,8 @@ pub mod bash;
 mod chat_completions;
 mod client;
 mod client_common;
+mod codebase_change_notice;
+mod codebase_snapshot;
 pub mod codex;
 mod codex_conversation;
 pub mod token_data;
@@ -29,9 +31,11 @@ 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;

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

@@ -0,0 +1,58 @@
+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,
+        } => {
+            determine_streamable_http_auth_status(
+                server_name,
+                url,
+                bearer_token_env_var.as_deref(),
+                store_mode,
+            )
+            .await
+        }
+    }
+}

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

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

codex-rs/core/src/mcp_connection_manager.rs

@@ -8,6 +8,7 @@
 
 use std::collections::HashMap;
 use std::collections::HashSet;
+use std::env;
 use std::ffi::OsString;
 use std::sync::Arc;
 use std::time::Duration;
@@ -16,6 +17,7 @@ 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;
@@ -125,9 +127,11 @@ impl McpClientAdapter {
         bearer_token: Option<String>,
         params: mcp_types::InitializeRequestParams,
         startup_timeout: Duration,
+        store_mode: OAuthCredentialsStoreMode,
     ) -> Result<Self> {
         let client = Arc::new(
-            RmcpClient::new_streamable_http_client(&server_name, &url, bearer_token).await?,
+            RmcpClient::new_streamable_http_client(&server_name, &url, bearer_token, store_mode)
+                .await?,
         );
         client.initialize(params, Some(startup_timeout)).await?;
         Ok(McpClientAdapter::Rmcp(client))
@@ -182,6 +186,7 @@ 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() {
@@ -202,9 +207,21 @@ impl McpConnectionManager {
                 continue;
             }
 
+            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 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 {
@@ -242,13 +259,14 @@ impl McpConnectionManager {
                         )
                         .await
                     }
-                    McpServerTransportConfig::StreamableHttp { url, bearer_token } => {
+                    McpServerTransportConfig::StreamableHttp { url, .. } => {
                         McpClientAdapter::new_streamable_http_client(
                             server_name.clone(),
                             url,
-                            bearer_token,
+                            resolved_bearer_token.unwrap_or_default(),
                             params,
                             startup_timeout,
+                            store_mode,
                         )
                         .await
                     }
@@ -336,6 +354,33 @@ 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.
 async fn list_all_tools(clients: &HashMap<String, ManagedClient>) -> Result<Vec<ToolInfo>> {

codex-rs/core/src/model_family.rs

@@ -119,8 +119,10 @@ pub fn find_family_for_model(mut slug: &str) -> Option<ModelFamily> {
             reasoning_summary_format: ReasoningSummaryFormat::Experimental,
             base_instructions: GPT_5_CODEX_INSTRUCTIONS.to_string(),
             experimental_supported_tools: vec![
+                "grep_files".to_string(),
+                "list_dir".to_string(),
                 "read_file".to_string(),
-                "test_sync_tool".to_string()
+                "test_sync_tool".to_string(),
             ],
             supports_parallel_tool_calls: true,
         )
@@ -133,7 +135,11 @@ pub fn find_family_for_model(mut slug: &str) -> Option<ModelFamily> {
             reasoning_summary_format: ReasoningSummaryFormat::Experimental,
             base_instructions: GPT_5_CODEX_INSTRUCTIONS.to_string(),
             apply_patch_tool_type: Some(ApplyPatchToolType::Freeform),
-            experimental_supported_tools: vec!["read_file".to_string()],
+            experimental_supported_tools: vec![
+                "grep_files".to_string(),
+                "list_dir".to_string(),
+                "read_file".to_string(),
+            ],
             supports_parallel_tool_calls: true,
         )
 

codex-rs/core/src/state/session.rs

@@ -2,6 +2,7 @@
 
 use codex_protocol::models::ResponseItem;
 
+use crate::codebase_snapshot::CodebaseSnapshot;
 use crate::conversation_history::ConversationHistory;
 use crate::protocol::RateLimitSnapshot;
 use crate::protocol::TokenUsage;
@@ -13,6 +14,7 @@ pub(crate) struct SessionState {
     pub(crate) history: ConversationHistory,
     pub(crate) token_info: Option<TokenUsageInfo>,
     pub(crate) latest_rate_limits: Option<RateLimitSnapshot>,
+    pub(crate) codebase_snapshot: Option<CodebaseSnapshot>,
 }
 
 impl SessionState {

codex-rs/core/src/state/turn.rs

@@ -34,6 +34,16 @@ pub(crate) enum TaskKind {
     Compact,
 }
 
+impl TaskKind {
+    pub(crate) fn header_value(self) -> &'static str {
+        match self {
+            TaskKind::Regular => "standard",
+            TaskKind::Review => "review",
+            TaskKind::Compact => "compact",
+        }
+    }
+}
+
 #[derive(Clone)]
 pub(crate) struct RunningTask {
     pub(crate) handle: AbortHandle,
@@ -113,3 +123,15 @@ impl ActiveTurn {
         }
     }
 }
+
+#[cfg(test)]
+mod tests {
+    use super::TaskKind;
+
+    #[test]
+    fn header_value_matches_expected_labels() {
+        assert_eq!(TaskKind::Regular.header_value(), "standard");
+        assert_eq!(TaskKind::Review.header_value(), "review");
+        assert_eq!(TaskKind::Compact.header_value(), "compact");
+    }
+}

codex-rs/core/src/tasks/regular.rs

@@ -27,6 +27,6 @@ impl SessionTask for RegularTask {
         input: Vec<InputItem>,
     ) -> Option<String> {
         let sess = session.clone_session();
-        run_task(sess, ctx, sub_id, input).await
+        run_task(sess, ctx, sub_id, input, TaskKind::Regular).await
     }
 }

codex-rs/core/src/tasks/review.rs

@@ -28,7 +28,7 @@ impl SessionTask for ReviewTask {
         input: Vec<InputItem>,
     ) -> Option<String> {
         let sess = session.clone_session();
-        run_task(sess, ctx, sub_id, input).await
+        run_task(sess, ctx, sub_id, input, TaskKind::Review).await
     }
 
     async fn abort(&self, session: Arc<SessionTaskContext>, sub_id: &str) {

codex-rs/core/src/tools/handlers/grep_files.rs

@@ -0,0 +1,272 @@
+use std::path::Path;
+use std::time::Duration;
+
+use async_trait::async_trait;
+use serde::Deserialize;
+use tokio::process::Command;
+use tokio::time::timeout;
+
+use crate::function_tool::FunctionCallError;
+use crate::tools::context::ToolInvocation;
+use crate::tools::context::ToolOutput;
+use crate::tools::context::ToolPayload;
+use crate::tools::registry::ToolHandler;
+use crate::tools::registry::ToolKind;
+
+pub struct GrepFilesHandler;
+
+const DEFAULT_LIMIT: usize = 100;
+const MAX_LIMIT: usize = 2000;
+const COMMAND_TIMEOUT: Duration = Duration::from_secs(30);
+
+fn default_limit() -> usize {
+    DEFAULT_LIMIT
+}
+
+#[derive(Deserialize)]
+struct GrepFilesArgs {
+    pattern: String,
+    #[serde(default)]
+    include: Option<String>,
+    #[serde(default)]
+    path: Option<String>,
+    #[serde(default = "default_limit")]
+    limit: usize,
+}
+
+#[async_trait]
+impl ToolHandler for GrepFilesHandler {
+    fn kind(&self) -> ToolKind {
+        ToolKind::Function
+    }
+
+    async fn handle(&self, invocation: ToolInvocation) -> Result<ToolOutput, FunctionCallError> {
+        let ToolInvocation { payload, turn, .. } = invocation;
+
+        let arguments = match payload {
+            ToolPayload::Function { arguments } => arguments,
+            _ => {
+                return Err(FunctionCallError::RespondToModel(
+                    "grep_files handler received unsupported payload".to_string(),
+                ));
+            }
+        };
+
+        let args: GrepFilesArgs = serde_json::from_str(&arguments).map_err(|err| {
+            FunctionCallError::RespondToModel(format!(
+                "failed to parse function arguments: {err:?}"
+            ))
+        })?;
+
+        let pattern = args.pattern.trim();
+        if pattern.is_empty() {
+            return Err(FunctionCallError::RespondToModel(
+                "pattern must not be empty".to_string(),
+            ));
+        }
+
+        if args.limit == 0 {
+            return Err(FunctionCallError::RespondToModel(
+                "limit must be greater than zero".to_string(),
+            ));
+        }
+
+        let limit = args.limit.min(MAX_LIMIT);
+        let search_path = turn.resolve_path(args.path.clone());
+
+        verify_path_exists(&search_path).await?;
+
+        let include = args.include.as_deref().map(str::trim).and_then(|val| {
+            if val.is_empty() {
+                None
+            } else {
+                Some(val.to_string())
+            }
+        });
+
+        let search_results =
+            run_rg_search(pattern, include.as_deref(), &search_path, limit, &turn.cwd).await?;
+
+        if search_results.is_empty() {
+            Ok(ToolOutput::Function {
+                content: "No matches found.".to_string(),
+                success: Some(false),
+            })
+        } else {
+            Ok(ToolOutput::Function {
+                content: search_results.join("\n"),
+                success: Some(true),
+            })
+        }
+    }
+}
+
+async fn verify_path_exists(path: &Path) -> Result<(), FunctionCallError> {
+    tokio::fs::metadata(path).await.map_err(|err| {
+        FunctionCallError::RespondToModel(format!("unable to access `{}`: {err}", path.display()))
+    })?;
+    Ok(())
+}
+
+async fn run_rg_search(
+    pattern: &str,
+    include: Option<&str>,
+    search_path: &Path,
+    limit: usize,
+    cwd: &Path,
+) -> Result<Vec<String>, FunctionCallError> {
+    let mut command = Command::new("rg");
+    command
+        .current_dir(cwd)
+        .arg("--files-with-matches")
+        .arg("--sortr=modified")
+        .arg("--regexp")
+        .arg(pattern)
+        .arg("--no-messages");
+
+    if let Some(glob) = include {
+        command.arg("--glob").arg(glob);
+    }
+
+    command.arg("--").arg(search_path);
+
+    let output = timeout(COMMAND_TIMEOUT, command.output())
+        .await
+        .map_err(|_| {
+            FunctionCallError::RespondToModel("rg timed out after 30 seconds".to_string())
+        })?
+        .map_err(|err| {
+            FunctionCallError::RespondToModel(format!(
+                "failed to launch rg: {err}. Ensure ripgrep is installed and on PATH."
+            ))
+        })?;
+
+    match output.status.code() {
+        Some(0) => Ok(parse_results(&output.stdout, limit)),
+        Some(1) => Ok(Vec::new()),
+        _ => {
+            let stderr = String::from_utf8_lossy(&output.stderr);
+            Err(FunctionCallError::RespondToModel(format!(
+                "rg failed: {stderr}"
+            )))
+        }
+    }
+}
+
+fn parse_results(stdout: &[u8], limit: usize) -> Vec<String> {
+    let mut results = Vec::new();
+    for line in stdout.split(|byte| *byte == b'\n') {
+        if line.is_empty() {
+            continue;
+        }
+        if let Ok(text) = std::str::from_utf8(line) {
+            if text.is_empty() {
+                continue;
+            }
+            results.push(text.to_string());
+            if results.len() == limit {
+                break;
+            }
+        }
+    }
+    results
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use std::process::Command as StdCommand;
+    use tempfile::tempdir;
+
+    #[test]
+    fn parses_basic_results() {
+        let stdout = b"/tmp/file_a.rs\n/tmp/file_b.rs\n";
+        let parsed = parse_results(stdout, 10);
+        assert_eq!(
+            parsed,
+            vec!["/tmp/file_a.rs".to_string(), "/tmp/file_b.rs".to_string()]
+        );
+    }
+
+    #[test]
+    fn parse_truncates_after_limit() {
+        let stdout = b"/tmp/file_a.rs\n/tmp/file_b.rs\n/tmp/file_c.rs\n";
+        let parsed = parse_results(stdout, 2);
+        assert_eq!(
+            parsed,
+            vec!["/tmp/file_a.rs".to_string(), "/tmp/file_b.rs".to_string()]
+        );
+    }
+
+    #[tokio::test]
+    async fn run_search_returns_results() -> anyhow::Result<()> {
+        if !rg_available() {
+            return Ok(());
+        }
+        let temp = tempdir().expect("create temp dir");
+        let dir = temp.path();
+        std::fs::write(dir.join("match_one.txt"), "alpha beta gamma").unwrap();
+        std::fs::write(dir.join("match_two.txt"), "alpha delta").unwrap();
+        std::fs::write(dir.join("other.txt"), "omega").unwrap();
+
+        let results = run_rg_search("alpha", None, dir, 10, dir).await?;
+        assert_eq!(results.len(), 2);
+        assert!(results.iter().any(|path| path.ends_with("match_one.txt")));
+        assert!(results.iter().any(|path| path.ends_with("match_two.txt")));
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn run_search_with_glob_filter() -> anyhow::Result<()> {
+        if !rg_available() {
+            return Ok(());
+        }
+        let temp = tempdir().expect("create temp dir");
+        let dir = temp.path();
+        std::fs::write(dir.join("match_one.rs"), "alpha beta gamma").unwrap();
+        std::fs::write(dir.join("match_two.txt"), "alpha delta").unwrap();
+
+        let results = run_rg_search("alpha", Some("*.rs"), dir, 10, dir).await?;
+        assert_eq!(results.len(), 1);
+        assert!(results.iter().all(|path| path.ends_with("match_one.rs")));
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn run_search_respects_limit() -> anyhow::Result<()> {
+        if !rg_available() {
+            return Ok(());
+        }
+        let temp = tempdir().expect("create temp dir");
+        let dir = temp.path();
+        std::fs::write(dir.join("one.txt"), "alpha one").unwrap();
+        std::fs::write(dir.join("two.txt"), "alpha two").unwrap();
+        std::fs::write(dir.join("three.txt"), "alpha three").unwrap();
+
+        let results = run_rg_search("alpha", None, dir, 2, dir).await?;
+        assert_eq!(results.len(), 2);
+        Ok(())
+    }
+
+    #[tokio::test]
+    async fn run_search_handles_no_matches() -> anyhow::Result<()> {
+        if !rg_available() {
+            return Ok(());
+        }
+        let temp = tempdir().expect("create temp dir");
+        let dir = temp.path();
+        std::fs::write(dir.join("one.txt"), "omega").unwrap();
+
+        let results = run_rg_search("alpha", None, dir, 5, dir).await?;
+        assert!(results.is_empty());
+        Ok(())
+    }
+
+    fn rg_available() -> bool {
+        StdCommand::new("rg")
+            .arg("--version")
+            .output()
+            .map(|output| output.status.success())
+            .unwrap_or(false)
+    }
+}

codex-rs/core/src/tools/handlers/list_dir.rs

@@ -0,0 +1,476 @@
+use std::collections::VecDeque;
+use std::ffi::OsStr;
+use std::fs::FileType;
+use std::path::Path;
+use std::path::PathBuf;
+
+use async_trait::async_trait;
+use codex_utils_string::take_bytes_at_char_boundary;
+use serde::Deserialize;
+use tokio::fs;
+
+use crate::function_tool::FunctionCallError;
+use crate::tools::context::ToolInvocation;
+use crate::tools::context::ToolOutput;
+use crate::tools::context::ToolPayload;
+use crate::tools::registry::ToolHandler;
+use crate::tools::registry::ToolKind;
+
+pub struct ListDirHandler;
+
+const MAX_ENTRY_LENGTH: usize = 500;
+const INDENTATION_SPACES: usize = 2;
+
+fn default_offset() -> usize {
+    1
+}
+
+fn default_limit() -> usize {
+    25
+}
+
+fn default_depth() -> usize {
+    2
+}
+
+#[derive(Deserialize)]
+struct ListDirArgs {
+    dir_path: String,
+    #[serde(default = "default_offset")]
+    offset: usize,
+    #[serde(default = "default_limit")]
+    limit: usize,
+    #[serde(default = "default_depth")]
+    depth: usize,
+}
+
+#[async_trait]
+impl ToolHandler for ListDirHandler {
+    fn kind(&self) -> ToolKind {
+        ToolKind::Function
+    }
+
+    async fn handle(&self, invocation: ToolInvocation) -> Result<ToolOutput, FunctionCallError> {
+        let ToolInvocation { payload, .. } = invocation;
+
+        let arguments = match payload {
+            ToolPayload::Function { arguments } => arguments,
+            _ => {
+                return Err(FunctionCallError::RespondToModel(
+                    "list_dir handler received unsupported payload".to_string(),
+                ));
+            }
+        };
+
+        let args: ListDirArgs = serde_json::from_str(&arguments).map_err(|err| {
+            FunctionCallError::RespondToModel(format!(
+                "failed to parse function arguments: {err:?}"
+            ))
+        })?;
+
+        let ListDirArgs {
+            dir_path,
+            offset,
+            limit,
+            depth,
+        } = args;
+
+        if offset == 0 {
+            return Err(FunctionCallError::RespondToModel(
+                "offset must be a 1-indexed entry number".to_string(),
+            ));
+        }
+
+        if limit == 0 {
+            return Err(FunctionCallError::RespondToModel(
+                "limit must be greater than zero".to_string(),
+            ));
+        }
+
+        if depth == 0 {
+            return Err(FunctionCallError::RespondToModel(
+                "depth must be greater than zero".to_string(),
+            ));
+        }
+
+        let path = PathBuf::from(&dir_path);
+        if !path.is_absolute() {
+            return Err(FunctionCallError::RespondToModel(
+                "dir_path must be an absolute path".to_string(),
+            ));
+        }
+
+        let entries = list_dir_slice(&path, offset, limit, depth).await?;
+        let mut output = Vec::with_capacity(entries.len() + 1);
+        output.push(format!("Absolute path: {}", path.display()));
+        output.extend(entries);
+        Ok(ToolOutput::Function {
+            content: output.join("\n"),
+            success: Some(true),
+        })
+    }
+}
+
+async fn list_dir_slice(
+    path: &Path,
+    offset: usize,
+    limit: usize,
+    depth: usize,
+) -> Result<Vec<String>, FunctionCallError> {
+    let mut entries = Vec::new();
+    collect_entries(path, Path::new(""), depth, &mut entries).await?;
+
+    if entries.is_empty() {
+        return Ok(Vec::new());
+    }
+
+    let start_index = offset - 1;
+    if start_index >= entries.len() {
+        return Err(FunctionCallError::RespondToModel(
+            "offset exceeds directory entry count".to_string(),
+        ));
+    }
+
+    let remaining_entries = entries.len() - start_index;
+    let capped_limit = limit.min(remaining_entries);
+    let end_index = start_index + capped_limit;
+    let mut selected_entries = entries[start_index..end_index].to_vec();
+    selected_entries.sort_unstable_by(|a, b| a.name.cmp(&b.name));
+    let mut formatted = Vec::with_capacity(selected_entries.len());
+
+    for entry in &selected_entries {
+        formatted.push(format_entry_line(entry));
+    }
+
+    if end_index < entries.len() {
+        formatted.push(format!("More than {capped_limit} entries found"));
+    }
+
+    Ok(formatted)
+}
+
+async fn collect_entries(
+    dir_path: &Path,
+    relative_prefix: &Path,
+    depth: usize,
+    entries: &mut Vec<DirEntry>,
+) -> Result<(), FunctionCallError> {
+    let mut queue = VecDeque::new();
+    queue.push_back((dir_path.to_path_buf(), relative_prefix.to_path_buf(), depth));
+
+    while let Some((current_dir, prefix, remaining_depth)) = queue.pop_front() {
+        let mut read_dir = fs::read_dir(&current_dir).await.map_err(|err| {
+            FunctionCallError::RespondToModel(format!("failed to read directory: {err}"))
+        })?;
+
+        let mut dir_entries = Vec::new();
+
+        while let Some(entry) = read_dir.next_entry().await.map_err(|err| {
+            FunctionCallError::RespondToModel(format!("failed to read directory: {err}"))
+        })? {
+            let file_type = entry.file_type().await.map_err(|err| {
+                FunctionCallError::RespondToModel(format!("failed to inspect entry: {err}"))
+            })?;
+
+            let file_name = entry.file_name();
+            let relative_path = if prefix.as_os_str().is_empty() {
+                PathBuf::from(&file_name)
+            } else {
+                prefix.join(&file_name)
+            };
+
+            let display_name = format_entry_component(&file_name);
+            let display_depth = prefix.components().count();
+            let sort_key = format_entry_name(&relative_path);
+            let kind = DirEntryKind::from(&file_type);
+            dir_entries.push((
+                entry.path(),
+                relative_path,
+                kind,
+                DirEntry {
+                    name: sort_key,
+                    display_name,
+                    depth: display_depth,
+                    kind,
+                },
+            ));
+        }
+
+        dir_entries.sort_unstable_by(|a, b| a.3.name.cmp(&b.3.name));
+
+        for (entry_path, relative_path, kind, dir_entry) in dir_entries {
+            if kind == DirEntryKind::Directory && remaining_depth > 1 {
+                queue.push_back((entry_path, relative_path, remaining_depth - 1));
+            }
+            entries.push(dir_entry);
+        }
+    }
+
+    Ok(())
+}
+
+fn format_entry_name(path: &Path) -> String {
+    let normalized = path.to_string_lossy().replace("\\", "/");
+    if normalized.len() > MAX_ENTRY_LENGTH {
+        take_bytes_at_char_boundary(&normalized, MAX_ENTRY_LENGTH).to_string()
+    } else {
+        normalized
+    }
+}
+
+fn format_entry_component(name: &OsStr) -> String {
+    let normalized = name.to_string_lossy();
+    if normalized.len() > MAX_ENTRY_LENGTH {
+        take_bytes_at_char_boundary(&normalized, MAX_ENTRY_LENGTH).to_string()
+    } else {
+        normalized.to_string()
+    }
+}
+
+fn format_entry_line(entry: &DirEntry) -> String {
+    let indent = " ".repeat(entry.depth * INDENTATION_SPACES);
+    let mut name = entry.display_name.clone();
+    match entry.kind {
+        DirEntryKind::Directory => name.push('/'),
+        DirEntryKind::Symlink => name.push('@'),
+        DirEntryKind::Other => name.push('?'),
+        DirEntryKind::File => {}
+    }
+    format!("{indent}{name}")
+}
+
+#[derive(Clone)]
+struct DirEntry {
+    name: String,
+    display_name: String,
+    depth: usize,
+    kind: DirEntryKind,
+}
+
+#[derive(Clone, Copy, PartialEq, Eq)]
+enum DirEntryKind {
+    Directory,
+    File,
+    Symlink,
+    Other,
+}
+
+impl From<&FileType> for DirEntryKind {
+    fn from(file_type: &FileType) -> Self {
+        if file_type.is_symlink() {
+            DirEntryKind::Symlink
+        } else if file_type.is_dir() {
+            DirEntryKind::Directory
+        } else if file_type.is_file() {
+            DirEntryKind::File
+        } else {
+            DirEntryKind::Other
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+    use tempfile::tempdir;
+
+    #[tokio::test]
+    async fn lists_directory_entries() {
+        let temp = tempdir().expect("create tempdir");
+        let dir_path = temp.path();
+
+        let sub_dir = dir_path.join("nested");
+        tokio::fs::create_dir(&sub_dir)
+            .await
+            .expect("create sub dir");
+
+        let deeper_dir = sub_dir.join("deeper");
+        tokio::fs::create_dir(&deeper_dir)
+            .await
+            .expect("create deeper dir");
+
+        tokio::fs::write(dir_path.join("entry.txt"), b"content")
+            .await
+            .expect("write file");
+        tokio::fs::write(sub_dir.join("child.txt"), b"child")
+            .await
+            .expect("write child");
+        tokio::fs::write(deeper_dir.join("grandchild.txt"), b"grandchild")
+            .await
+            .expect("write grandchild");
+
+        #[cfg(unix)]
+        {
+            use std::os::unix::fs::symlink;
+            let link_path = dir_path.join("link");
+            symlink(dir_path.join("entry.txt"), &link_path).expect("create symlink");
+        }
+
+        let entries = list_dir_slice(dir_path, 1, 20, 3)
+            .await
+            .expect("list directory");
+
+        #[cfg(unix)]
+        let expected = vec![
+            "entry.txt".to_string(),
+            "link@".to_string(),
+            "nested/".to_string(),
+            "  child.txt".to_string(),
+            "  deeper/".to_string(),
+            "    grandchild.txt".to_string(),
+        ];
+
+        #[cfg(not(unix))]
+        let expected = vec![
+            "entry.txt".to_string(),
+            "nested/".to_string(),
+            "  child.txt".to_string(),
+            "  deeper/".to_string(),
+            "    grandchild.txt".to_string(),
+        ];
+
+        assert_eq!(entries, expected);
+    }
+
+    #[tokio::test]
+    async fn errors_when_offset_exceeds_entries() {
+        let temp = tempdir().expect("create tempdir");
+        let dir_path = temp.path();
+        tokio::fs::create_dir(dir_path.join("nested"))
+            .await
+            .expect("create sub dir");
+
+        let err = list_dir_slice(dir_path, 10, 1, 2)
+            .await
+            .expect_err("offset exceeds entries");
+        assert_eq!(
+            err,
+            FunctionCallError::RespondToModel("offset exceeds directory entry count".to_string())
+        );
+    }
+
+    #[tokio::test]
+    async fn respects_depth_parameter() {
+        let temp = tempdir().expect("create tempdir");
+        let dir_path = temp.path();
+        let nested = dir_path.join("nested");
+        let deeper = nested.join("deeper");
+        tokio::fs::create_dir(&nested).await.expect("create nested");
+        tokio::fs::create_dir(&deeper).await.expect("create deeper");
+        tokio::fs::write(dir_path.join("root.txt"), b"root")
+            .await
+            .expect("write root");
+        tokio::fs::write(nested.join("child.txt"), b"child")
+            .await
+            .expect("write nested");
+        tokio::fs::write(deeper.join("grandchild.txt"), b"deep")
+            .await
+            .expect("write deeper");
+
+        let entries_depth_one = list_dir_slice(dir_path, 1, 10, 1)
+            .await
+            .expect("list depth 1");
+        assert_eq!(
+            entries_depth_one,
+            vec!["nested/".to_string(), "root.txt".to_string(),]
+        );
+
+        let entries_depth_two = list_dir_slice(dir_path, 1, 20, 2)
+            .await
+            .expect("list depth 2");
+        assert_eq!(
+            entries_depth_two,
+            vec![
+                "nested/".to_string(),
+                "  child.txt".to_string(),
+                "  deeper/".to_string(),
+                "root.txt".to_string(),
+            ]
+        );
+
+        let entries_depth_three = list_dir_slice(dir_path, 1, 30, 3)
+            .await
+            .expect("list depth 3");
+        assert_eq!(
+            entries_depth_three,
+            vec![
+                "nested/".to_string(),
+                "  child.txt".to_string(),
+                "  deeper/".to_string(),
+                "    grandchild.txt".to_string(),
+                "root.txt".to_string(),
+            ]
+        );
+    }
+
+    #[tokio::test]
+    async fn handles_large_limit_without_overflow() {
+        let temp = tempdir().expect("create tempdir");
+        let dir_path = temp.path();
+        tokio::fs::write(dir_path.join("alpha.txt"), b"alpha")
+            .await
+            .expect("write alpha");
+        tokio::fs::write(dir_path.join("beta.txt"), b"beta")
+            .await
+            .expect("write beta");
+        tokio::fs::write(dir_path.join("gamma.txt"), b"gamma")
+            .await
+            .expect("write gamma");
+
+        let entries = list_dir_slice(dir_path, 2, usize::MAX, 1)
+            .await
+            .expect("list without overflow");
+        assert_eq!(
+            entries,
+            vec!["beta.txt".to_string(), "gamma.txt".to_string(),]
+        );
+    }
+
+    #[tokio::test]
+    async fn indicates_truncated_results() {
+        let temp = tempdir().expect("create tempdir");
+        let dir_path = temp.path();
+
+        for idx in 0..40 {
+            let file = dir_path.join(format!("file_{idx:02}.txt"));
+            tokio::fs::write(file, b"content")
+                .await
+                .expect("write file");
+        }
+
+        let entries = list_dir_slice(dir_path, 1, 25, 1)
+            .await
+            .expect("list directory");
+        assert_eq!(entries.len(), 26);
+        assert_eq!(
+            entries.last(),
+            Some(&"More than 25 entries found".to_string())
+        );
+    }
+
+    #[tokio::test]
+    async fn bfs_truncation() -> anyhow::Result<()> {
+        let temp = tempdir()?;
+        let dir_path = temp.path();
+        let nested = dir_path.join("nested");
+        let deeper = nested.join("deeper");
+        tokio::fs::create_dir(&nested).await?;
+        tokio::fs::create_dir(&deeper).await?;
+        tokio::fs::write(dir_path.join("root.txt"), b"root").await?;
+        tokio::fs::write(nested.join("child.txt"), b"child").await?;
+        tokio::fs::write(deeper.join("grandchild.txt"), b"deep").await?;
+
+        let entries_depth_three = list_dir_slice(dir_path, 1, 3, 3).await?;
+        assert_eq!(
+            entries_depth_three,
+            vec![
+                "nested/".to_string(),
+                "  child.txt".to_string(),
+                "root.txt".to_string(),
+                "More than 3 entries found".to_string()
+            ]
+        );
+
+        Ok(())
+    }
+}

codex-rs/core/src/tools/handlers/mod.rs

@@ -1,5 +1,7 @@
 pub mod apply_patch;
 mod exec_stream;
+mod grep_files;
+mod list_dir;
 mod mcp;
 mod plan;
 mod read_file;
@@ -12,6 +14,8 @@ pub use plan::PLAN_TOOL;
 
 pub use apply_patch::ApplyPatchHandler;
 pub use exec_stream::ExecStreamHandler;
+pub use grep_files::GrepFilesHandler;
+pub use list_dir::ListDirHandler;
 pub use mcp::McpHandler;
 pub use plan::PlanHandler;
 pub use read_file::ReadFileHandler;

codex-rs/core/src/tools/parallel.rs

@@ -1,6 +1,8 @@
 use std::sync::Arc;
 
-use tokio::task::JoinHandle;
+use tokio::sync::RwLock;
+use tokio_util::either::Either;
+use tokio_util::task::AbortOnDropHandle;
 
 use crate::codex::Session;
 use crate::codex::TurnContext;
@@ -11,20 +13,13 @@ use crate::tools::router::ToolCall;
 use crate::tools::router::ToolRouter;
 use codex_protocol::models::ResponseInputItem;
 
-use crate::codex::ProcessedResponseItem;
-
-struct PendingToolCall {
-    index: usize,
-    handle: JoinHandle<Result<ResponseInputItem, FunctionCallError>>,
-}
-
 pub(crate) struct ToolCallRuntime {
     router: Arc<ToolRouter>,
     session: Arc<Session>,
     turn_context: Arc<TurnContext>,
     tracker: SharedTurnDiffTracker,
     sub_id: String,
-    pending_calls: Vec<PendingToolCall>,
+    parallel_execution: Arc<RwLock<()>>,
 }
 
 impl ToolCallRuntime {
@@ -41,97 +36,45 @@ impl ToolCallRuntime {
             turn_context,
             tracker,
             sub_id,
-            pending_calls: Vec::new(),
+            parallel_execution: Arc::new(RwLock::new(())),
         }
     }
 
-    pub(crate) async fn handle_tool_call(
-        &mut self,
+    pub(crate) fn handle_tool_call(
+        &self,
         call: ToolCall,
-        output_index: usize,
-        output: &mut [ProcessedResponseItem],
-    ) -> Result<(), CodexErr> {
+    ) -> impl std::future::Future<Output = Result<ResponseInputItem, CodexErr>> {
         let supports_parallel = self.router.tool_supports_parallel(&call.tool_name);
-        if supports_parallel {
-            self.spawn_parallel(call, output_index);
-        } else {
-            self.resolve_pending(output).await?;
-            let response = self.dispatch_serial(call).await?;
-            let slot = output.get_mut(output_index).ok_or_else(|| {
-                CodexErr::Fatal(format!("tool output index {output_index} out of bounds"))
-            })?;
-            slot.response = Some(response);
-        }
 
-        Ok(())
-    }
-
-    pub(crate) fn abort_all(&mut self) {
-        while let Some(pending) = self.pending_calls.pop() {
-            pending.handle.abort();
-        }
-    }
-
-    pub(crate) async fn resolve_pending(
-        &mut self,
-        output: &mut [ProcessedResponseItem],
-    ) -> Result<(), CodexErr> {
-        while let Some(PendingToolCall { index, handle }) = self.pending_calls.pop() {
-            match handle.await {
-                Ok(Ok(response)) => {
-                    if let Some(slot) = output.get_mut(index) {
-                        slot.response = Some(response);
-                    }
-                }
-                Ok(Err(FunctionCallError::Fatal(message))) => {
-                    self.abort_all();
-                    return Err(CodexErr::Fatal(message));
-                }
-                Ok(Err(other)) => {
-                    self.abort_all();
-                    return Err(CodexErr::Fatal(other.to_string()));
-                }
-                Err(join_err) => {
-                    self.abort_all();
-                    return Err(CodexErr::Fatal(format!(
-                        "tool task failed to join: {join_err}"
-                    )));
-                }
-            }
-        }
-
-        Ok(())
-    }
-
-    fn spawn_parallel(&mut self, call: ToolCall, index: usize) {
         let router = Arc::clone(&self.router);
         let session = Arc::clone(&self.session);
         let turn = Arc::clone(&self.turn_context);
         let tracker = Arc::clone(&self.tracker);
         let sub_id = self.sub_id.clone();
-        let handle = tokio::spawn(async move {
-            router
-                .dispatch_tool_call(session, turn, tracker, sub_id, call)
-                .await
-        });
-        self.pending_calls.push(PendingToolCall { index, handle });
-    }
+        let lock = Arc::clone(&self.parallel_execution);
 
-    async fn dispatch_serial(&self, call: ToolCall) -> Result<ResponseInputItem, CodexErr> {
-        match self
-            .router
-            .dispatch_tool_call(
-                Arc::clone(&self.session),
-                Arc::clone(&self.turn_context),
-                Arc::clone(&self.tracker),
-                self.sub_id.clone(),
-                call,
-            )
-            .await
-        {
-            Ok(response) => Ok(response),
-            Err(FunctionCallError::Fatal(message)) => Err(CodexErr::Fatal(message)),
-            Err(other) => Err(CodexErr::Fatal(other.to_string())),
+        let handle: AbortOnDropHandle<Result<ResponseInputItem, FunctionCallError>> =
+            AbortOnDropHandle::new(tokio::spawn(async move {
+                let _guard = if supports_parallel {
+                    Either::Left(lock.read().await)
+                } else {
+                    Either::Right(lock.write().await)
+                };
+
+                router
+                    .dispatch_tool_call(session, turn, tracker, sub_id, call)
+                    .await
+            }));
+
+        async move {
+            match handle.await {
+                Ok(Ok(response)) => Ok(response),
+                Ok(Err(FunctionCallError::Fatal(message))) => Err(CodexErr::Fatal(message)),
+                Ok(Err(other)) => Err(CodexErr::Fatal(other.to_string())),
+                Err(err) => Err(CodexErr::Fatal(format!(
+                    "tool task failed to receive: {err:?}"
+                ))),
+            }
         }
     }
 }

codex-rs/core/src/tools/spec.rs

@@ -1,5 +1,7 @@
 use crate::client_common::tools::ResponsesApiTool;
 use crate::client_common::tools::ToolSpec;
+use crate::features::Feature;
+use crate::features::Features;
 use crate::model_family::ModelFamily;
 use crate::tools::handlers::PLAN_TOOL;
 use crate::tools::handlers::apply_patch::ApplyPatchToolType;
@@ -33,26 +35,23 @@ pub(crate) struct ToolsConfig {
 
 pub(crate) struct ToolsConfigParams<'a> {
     pub(crate) model_family: &'a ModelFamily,
-    pub(crate) include_plan_tool: bool,
-    pub(crate) include_apply_patch_tool: bool,
-    pub(crate) include_web_search_request: bool,
-    pub(crate) use_streamable_shell_tool: bool,
-    pub(crate) include_view_image_tool: bool,
-    pub(crate) experimental_unified_exec_tool: bool,
+    pub(crate) features: &'a Features,
 }
 
 impl ToolsConfig {
     pub fn new(params: &ToolsConfigParams) -> Self {
         let ToolsConfigParams {
             model_family,
-            include_plan_tool,
-            include_apply_patch_tool,
-            include_web_search_request,
-            use_streamable_shell_tool,
-            include_view_image_tool,
-            experimental_unified_exec_tool,
+            features,
         } = params;
-        let shell_type = if *use_streamable_shell_tool {
+        let use_streamable_shell_tool = features.enabled(Feature::StreamableShell);
+        let experimental_unified_exec_tool = features.enabled(Feature::UnifiedExec);
+        let include_plan_tool = features.enabled(Feature::PlanTool);
+        let include_apply_patch_tool = features.enabled(Feature::ApplyPatchFreeform);
+        let include_web_search_request = features.enabled(Feature::WebSearchRequest);
+        let include_view_image_tool = features.enabled(Feature::ViewImageTool);
+
+        let shell_type = if use_streamable_shell_tool {
             ConfigShellToolType::Streamable
         } else if model_family.uses_local_shell_tool {
             ConfigShellToolType::Local
@@ -64,7 +63,7 @@ impl ToolsConfig {
             Some(ApplyPatchToolType::Freeform) => Some(ApplyPatchToolType::Freeform),
             Some(ApplyPatchToolType::Function) => Some(ApplyPatchToolType::Function),
             None => {
-                if *include_apply_patch_tool {
+                if include_apply_patch_tool {
                     Some(ApplyPatchToolType::Freeform)
                 } else {
                     None
@@ -74,11 +73,11 @@ impl ToolsConfig {
 
         Self {
             shell_type,
-            plan_tool: *include_plan_tool,
+            plan_tool: include_plan_tool,
             apply_patch_tool_type,
-            web_search_request: *include_web_search_request,
-            include_view_image_tool: *include_view_image_tool,
-            experimental_unified_exec_tool: *experimental_unified_exec_tool,
+            web_search_request: include_web_search_request,
+            include_view_image_tool,
+            experimental_unified_exec_tool,
             experimental_supported_tools: model_family.experimental_supported_tools.clone(),
         }
     }
@@ -320,6 +319,56 @@ fn create_test_sync_tool() -> ToolSpec {
     })
 }
 
+fn create_grep_files_tool() -> ToolSpec {
+    let mut properties = BTreeMap::new();
+    properties.insert(
+        "pattern".to_string(),
+        JsonSchema::String {
+            description: Some("Regular expression pattern to search for.".to_string()),
+        },
+    );
+    properties.insert(
+        "include".to_string(),
+        JsonSchema::String {
+            description: Some(
+                "Optional glob that limits which files are searched (e.g. \"*.rs\" or \
+                 \"*.{ts,tsx}\")."
+                    .to_string(),
+            ),
+        },
+    );
+    properties.insert(
+        "path".to_string(),
+        JsonSchema::String {
+            description: Some(
+                "Directory or file path to search. Defaults to the session's working directory."
+                    .to_string(),
+            ),
+        },
+    );
+    properties.insert(
+        "limit".to_string(),
+        JsonSchema::Number {
+            description: Some(
+                "Maximum number of file paths to return (defaults to 100).".to_string(),
+            ),
+        },
+    );
+
+    ToolSpec::Function(ResponsesApiTool {
+        name: "grep_files".to_string(),
+        description: "Finds files whose contents match the pattern and lists them by modification \
+                      time."
+            .to_string(),
+        strict: false,
+        parameters: JsonSchema::Object {
+            properties,
+            required: Some(vec!["pattern".to_string()]),
+            additional_properties: Some(false.into()),
+        },
+    })
+}
+
 fn create_read_file_tool() -> ToolSpec {
     let mut properties = BTreeMap::new();
     properties.insert(
@@ -342,11 +391,72 @@ fn create_read_file_tool() -> ToolSpec {
             description: Some("The maximum number of lines to return.".to_string()),
         },
     );
+    properties.insert(
+        "mode".to_string(),
+        JsonSchema::String {
+            description: Some(
+                "Optional mode selector: \"slice\" for simple ranges (default) or \"indentation\" \
+                 to expand around an anchor line."
+                    .to_string(),
+            ),
+        },
+    );
+
+    let mut indentation_properties = BTreeMap::new();
+    indentation_properties.insert(
+        "anchor_line".to_string(),
+        JsonSchema::Number {
+            description: Some(
+                "Anchor line to center the indentation lookup on (defaults to offset).".to_string(),
+            ),
+        },
+    );
+    indentation_properties.insert(
+        "max_levels".to_string(),
+        JsonSchema::Number {
+            description: Some(
+                "How many parent indentation levels (smaller indents) to include.".to_string(),
+            ),
+        },
+    );
+    indentation_properties.insert(
+        "include_siblings".to_string(),
+        JsonSchema::Boolean {
+            description: Some(
+                "When true, include additional blocks that share the anchor indentation."
+                    .to_string(),
+            ),
+        },
+    );
+    indentation_properties.insert(
+        "include_header".to_string(),
+        JsonSchema::Boolean {
+            description: Some(
+                "Include doc comments or attributes directly above the selected block.".to_string(),
+            ),
+        },
+    );
+    indentation_properties.insert(
+        "max_lines".to_string(),
+        JsonSchema::Number {
+            description: Some(
+                "Hard cap on the number of lines returned when using indentation mode.".to_string(),
+            ),
+        },
+    );
+    properties.insert(
+        "indentation".to_string(),
+        JsonSchema::Object {
+            properties: indentation_properties,
+            required: None,
+            additional_properties: Some(false.into()),
+        },
+    );
 
     ToolSpec::Function(ResponsesApiTool {
         name: "read_file".to_string(),
         description:
-            "Reads a local file with 1-indexed line numbers and returns up to the requested number of lines."
+            "Reads a local file with 1-indexed line numbers, supporting slice and indentation-aware block modes."
                 .to_string(),
         strict: false,
         parameters: JsonSchema::Object {
@@ -356,6 +466,51 @@ fn create_read_file_tool() -> ToolSpec {
         },
     })
 }
+
+fn create_list_dir_tool() -> ToolSpec {
+    let mut properties = BTreeMap::new();
+    properties.insert(
+        "dir_path".to_string(),
+        JsonSchema::String {
+            description: Some("Absolute path to the directory to list.".to_string()),
+        },
+    );
+    properties.insert(
+        "offset".to_string(),
+        JsonSchema::Number {
+            description: Some(
+                "The entry number to start listing from. Must be 1 or greater.".to_string(),
+            ),
+        },
+    );
+    properties.insert(
+        "limit".to_string(),
+        JsonSchema::Number {
+            description: Some("The maximum number of entries to return.".to_string()),
+        },
+    );
+    properties.insert(
+        "depth".to_string(),
+        JsonSchema::Number {
+            description: Some(
+                "The maximum directory depth to traverse. Must be 1 or greater.".to_string(),
+            ),
+        },
+    );
+
+    ToolSpec::Function(ResponsesApiTool {
+        name: "list_dir".to_string(),
+        description:
+            "Lists entries in a local directory with 1-indexed entry numbers and simple type labels."
+                .to_string(),
+        strict: false,
+        parameters: JsonSchema::Object {
+            properties,
+            required: Some(vec!["dir_path".to_string()]),
+            additional_properties: Some(false.into()),
+        },
+    })
+}
 /// TODO(dylan): deprecate once we get rid of json tool
 #[derive(Serialize, Deserialize)]
 pub(crate) struct ApplyPatchToolArgs {
@@ -565,6 +720,8 @@ pub(crate) fn build_specs(
     use crate::exec_command::create_write_stdin_tool_for_responses_api;
     use crate::tools::handlers::ApplyPatchHandler;
     use crate::tools::handlers::ExecStreamHandler;
+    use crate::tools::handlers::GrepFilesHandler;
+    use crate::tools::handlers::ListDirHandler;
     use crate::tools::handlers::McpHandler;
     use crate::tools::handlers::PlanHandler;
     use crate::tools::handlers::ReadFileHandler;
@@ -632,8 +789,16 @@ pub(crate) fn build_specs(
 
     if config
         .experimental_supported_tools
-        .iter()
-        .any(|tool| tool == "read_file")
+        .contains(&"grep_files".to_string())
+    {
+        let grep_files_handler = Arc::new(GrepFilesHandler);
+        builder.push_spec_with_parallel_support(create_grep_files_tool(), true);
+        builder.register_handler("grep_files", grep_files_handler);
+    }
+
+    if config
+        .experimental_supported_tools
+        .contains(&"read_file".to_string())
     {
         let read_file_handler = Arc::new(ReadFileHandler);
         builder.push_spec_with_parallel_support(create_read_file_tool(), true);
@@ -643,7 +808,16 @@ pub(crate) fn build_specs(
     if config
         .experimental_supported_tools
         .iter()
-        .any(|tool| tool == "test_sync_tool")
+        .any(|tool| tool == "list_dir")
+    {
+        let list_dir_handler = Arc::new(ListDirHandler);
+        builder.push_spec_with_parallel_support(create_list_dir_tool(), true);
+        builder.register_handler("list_dir", list_dir_handler);
+    }
+
+    if config
+        .experimental_supported_tools
+        .contains(&"test_sync_tool".to_string())
     {
         let test_sync_handler = Arc::new(TestSyncHandler);
         builder.push_spec_with_parallel_support(create_test_sync_tool(), true);
@@ -731,14 +905,13 @@ mod tests {
     fn test_build_specs() {
         let model_family = find_family_for_model("codex-mini-latest")
             .expect("codex-mini-latest should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::PlanTool);
+        features.enable(Feature::WebSearchRequest);
+        features.enable(Feature::UnifiedExec);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: true,
-            include_apply_patch_tool: false,
-            include_web_search_request: true,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
         let (tools, _) = build_specs(&config, Some(HashMap::new())).build();
 
@@ -751,14 +924,13 @@ mod tests {
     #[test]
     fn test_build_specs_default_shell() {
         let model_family = find_family_for_model("o3").expect("o3 should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::PlanTool);
+        features.enable(Feature::WebSearchRequest);
+        features.enable(Feature::UnifiedExec);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: true,
-            include_apply_patch_tool: false,
-            include_web_search_request: true,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
         let (tools, _) = build_specs(&config, Some(HashMap::new())).build();
 
@@ -773,18 +945,18 @@ mod tests {
     fn test_parallel_support_flags() {
         let model_family = find_family_for_model("gpt-5-codex")
             .expect("codex-mini-latest should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.disable(Feature::ViewImageTool);
+        features.enable(Feature::UnifiedExec);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: false,
-            include_web_search_request: false,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: false,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
         let (tools, _) = build_specs(&config, None).build();
 
         assert!(!find_tool(&tools, "unified_exec").supports_parallel_tool_calls);
+        assert!(find_tool(&tools, "grep_files").supports_parallel_tool_calls);
+        assert!(find_tool(&tools, "list_dir").supports_parallel_tool_calls);
         assert!(find_tool(&tools, "read_file").supports_parallel_tool_calls);
     }
 
@@ -792,14 +964,11 @@ mod tests {
     fn test_test_model_family_includes_sync_tool() {
         let model_family = find_family_for_model("test-gpt-5-codex")
             .expect("test-gpt-5-codex should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.disable(Feature::ViewImageTool);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: false,
-            include_web_search_request: false,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: false,
-            experimental_unified_exec_tool: false,
+            features: &features,
         });
         let (tools, _) = build_specs(&config, None).build();
 
@@ -813,19 +982,23 @@ mod tests {
                 .iter()
                 .any(|tool| tool_name(&tool.spec) == "read_file")
         );
+        assert!(
+            tools
+                .iter()
+                .any(|tool| tool_name(&tool.spec) == "grep_files")
+        );
+        assert!(tools.iter().any(|tool| tool_name(&tool.spec) == "list_dir"));
     }
 
     #[test]
     fn test_build_specs_mcp_tools() {
         let model_family = find_family_for_model("o3").expect("o3 should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::UnifiedExec);
+        features.enable(Feature::WebSearchRequest);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: false,
-            include_web_search_request: true,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
         let (tools, _) = build_specs(
             &config,
@@ -923,14 +1096,11 @@ mod tests {
     #[test]
     fn test_build_specs_mcp_tools_sorted_by_name() {
         let model_family = find_family_for_model("o3").expect("o3 should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::UnifiedExec);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: false,
-            include_web_search_request: false,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
 
         // Intentionally construct a map with keys that would sort alphabetically.
@@ -1000,14 +1170,12 @@ mod tests {
     fn test_mcp_tool_property_missing_type_defaults_to_string() {
         let model_family = find_family_for_model("gpt-5-codex")
             .expect("gpt-5-codex should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::UnifiedExec);
+        features.enable(Feature::WebSearchRequest);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: false,
-            include_web_search_request: true,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
 
         let (tools, _) = build_specs(
@@ -1069,14 +1237,12 @@ mod tests {
     fn test_mcp_tool_integer_normalized_to_number() {
         let model_family = find_family_for_model("gpt-5-codex")
             .expect("gpt-5-codex should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::UnifiedExec);
+        features.enable(Feature::WebSearchRequest);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: false,
-            include_web_search_request: true,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
 
         let (tools, _) = build_specs(
@@ -1133,14 +1299,13 @@ mod tests {
     fn test_mcp_tool_array_without_items_gets_default_string_items() {
         let model_family = find_family_for_model("gpt-5-codex")
             .expect("gpt-5-codex should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::UnifiedExec);
+        features.enable(Feature::WebSearchRequest);
+        features.enable(Feature::ApplyPatchFreeform);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: true,
-            include_web_search_request: true,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
 
         let (tools, _) = build_specs(
@@ -1200,14 +1365,12 @@ mod tests {
     fn test_mcp_tool_anyof_defaults_to_string() {
         let model_family = find_family_for_model("gpt-5-codex")
             .expect("gpt-5-codex should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::UnifiedExec);
+        features.enable(Feature::WebSearchRequest);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: false,
-            include_web_search_request: true,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
 
         let (tools, _) = build_specs(
@@ -1279,14 +1442,12 @@ mod tests {
     fn test_get_openai_tools_mcp_tools_with_additional_properties_schema() {
         let model_family = find_family_for_model("gpt-5-codex")
             .expect("gpt-5-codex should be a valid model family");
+        let mut features = Features::with_defaults();
+        features.enable(Feature::UnifiedExec);
+        features.enable(Feature::WebSearchRequest);
         let config = ToolsConfig::new(&ToolsConfigParams {
             model_family: &model_family,
-            include_plan_tool: false,
-            include_apply_patch_tool: false,
-            include_web_search_request: true,
-            use_streamable_shell_tool: false,
-            include_view_image_tool: true,
-            experimental_unified_exec_tool: true,
+            features: &features,
         });
         let (tools, _) = build_specs(
             &config,

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

@@ -110,11 +110,22 @@ impl ManagedUnifiedExecSession {
         let buffer_clone = Arc::clone(&output_buffer);
         let notify_clone = Arc::clone(&output_notify);
         let output_task = tokio::spawn(async move {
-            while let Ok(chunk) = receiver.recv().await {
-                let mut guard = buffer_clone.lock().await;
-                guard.push_chunk(chunk);
-                drop(guard);
-                notify_clone.notify_waiters();
+            loop {
+                match receiver.recv().await {
+                    Ok(chunk) => {
+                        let mut guard = buffer_clone.lock().await;
+                        guard.push_chunk(chunk);
+                        drop(guard);
+                        notify_clone.notify_waiters();
+                    }
+                    // If we lag behind the broadcast buffer, skip missed
+                    // messages but keep the task alive to continue streaming.
+                    Err(tokio::sync::broadcast::error::RecvError::Lagged(_)) => {
+                        continue;
+                    }
+                    // When the sender closes, exit the task.
+                    Err(tokio::sync::broadcast::error::RecvError::Closed) => break,
+                }
             }
         });
 

codex-rs/core/src/user_notification.rs

@@ -49,6 +49,7 @@ impl UserNotifier {
 pub(crate) enum UserNotification {
     #[serde(rename_all = "kebab-case")]
     AgentTurnComplete {
+        thread_id: String,
         turn_id: String,
 
         /// Messages that the user sent to the agent to initiate the turn.
@@ -67,6 +68,7 @@ mod tests {
     #[test]
     fn test_user_notification() -> Result<()> {
         let notification = UserNotification::AgentTurnComplete {
+            thread_id: "b5f6c1c2-1111-2222-3333-444455556666".to_string(),
             turn_id: "12345".to_string(),
             input_messages: vec!["Rename `foo` to `bar` and update the callsites.".to_string()],
             last_assistant_message: Some(
@@ -76,7 +78,7 @@ mod tests {
         let serialized = serde_json::to_string(&notification)?;
         assert_eq!(
             serialized,
-            r#"{"type":"agent-turn-complete","turn-id":"12345","input-messages":["Rename `foo` to `bar` and update the callsites."],"last-assistant-message":"Rename complete and verified `cargo build` succeeds."}"#
+            r#"{"type":"agent-turn-complete","thread-id":"b5f6c1c2-1111-2222-3333-444455556666","turn-id":"12345","input-messages":["Rename `foo` to `bar` and update the callsites."],"last-assistant-message":"Rename complete and verified `cargo build` succeeds."}"#
         );
         Ok(())
     }

codex-rs/core/tests/chat_completions_sse.rs

@@ -1,3 +1,4 @@
+use assert_matches::assert_matches;
 use std::sync::Arc;
 use tracing_test::traced_test;
 
@@ -178,7 +179,7 @@ async fn streams_text_without_reasoning() {
         other => panic!("expected terminal message, got {other:?}"),
     }
 
-    assert!(matches!(events[2], ResponseEvent::Completed { .. }));
+    assert_matches!(events[2], ResponseEvent::Completed { .. });
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -219,7 +220,7 @@ async fn streams_reasoning_from_string_delta() {
         other => panic!("expected message item, got {other:?}"),
     }
 
-    assert!(matches!(events[4], ResponseEvent::Completed { .. }));
+    assert_matches!(events[4], ResponseEvent::Completed { .. });
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -266,7 +267,7 @@ async fn streams_reasoning_from_object_delta() {
         other => panic!("expected message item, got {other:?}"),
     }
 
-    assert!(matches!(events[5], ResponseEvent::Completed { .. }));
+    assert_matches!(events[5], ResponseEvent::Completed { .. });
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -293,7 +294,7 @@ async fn streams_reasoning_from_final_message() {
         other => panic!("expected reasoning item, got {other:?}"),
     }
 
-    assert!(matches!(events[2], ResponseEvent::Completed { .. }));
+    assert_matches!(events[2], ResponseEvent::Completed { .. });
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -337,7 +338,7 @@ async fn streams_reasoning_before_tool_call() {
         other => panic!("expected function call, got {other:?}"),
     }
 
-    assert!(matches!(events[3], ResponseEvent::Completed { .. }));
+    assert_matches!(events[3], ResponseEvent::Completed { .. });
 }
 
 #[tokio::test]

codex-rs/core/tests/common/Cargo.toml

@@ -10,7 +10,10 @@ path = "lib.rs"
 anyhow = { workspace = true }
 assert_cmd = { workspace = true }
 codex-core = { workspace = true }
+notify = { workspace = true }
+regex-lite = { workspace = true }
 serde_json = { workspace = true }
 tempfile = { workspace = true }
 tokio = { workspace = true, features = ["time"] }
+walkdir = { workspace = true }
 wiremock = { workspace = true }

codex-rs/core/tests/common/lib.rs

@@ -6,6 +6,7 @@ use codex_core::CodexConversation;
 use codex_core::config::Config;
 use codex_core::config::ConfigOverrides;
 use codex_core::config::ConfigToml;
+use regex_lite::Regex;
 
 #[cfg(target_os = "linux")]
 use assert_cmd::cargo::cargo_bin;
@@ -14,6 +15,16 @@ pub mod responses;
 pub mod test_codex;
 pub mod test_codex_exec;
 
+#[track_caller]
+pub fn assert_regex_match<'s>(pattern: &str, actual: &'s str) -> regex_lite::Captures<'s> {
+    let regex = Regex::new(pattern).unwrap_or_else(|err| {
+        panic!("failed to compile regex {pattern:?}: {err}");
+    });
+    regex
+        .captures(actual)
+        .unwrap_or_else(|| panic!("regex {pattern:?} did not match {actual:?}"))
+}
+
 /// Returns a default `Config` whose on-disk state is confined to the provided
 /// temporary directory. Using a per-test directory keeps tests hermetic and
 /// avoids clobbering a developer’s real `~/.codex`.
@@ -153,6 +164,149 @@ pub fn sandbox_network_env_var() -> &'static str {
     codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR
 }
 
+pub mod fs_wait {
+    use anyhow::Result;
+    use anyhow::anyhow;
+    use notify::RecursiveMode;
+    use notify::Watcher;
+    use std::path::Path;
+    use std::path::PathBuf;
+    use std::sync::mpsc;
+    use std::sync::mpsc::RecvTimeoutError;
+    use std::time::Duration;
+    use std::time::Instant;
+    use tokio::task;
+    use walkdir::WalkDir;
+
+    pub async fn wait_for_path_exists(
+        path: impl Into<PathBuf>,
+        timeout: Duration,
+    ) -> Result<PathBuf> {
+        let path = path.into();
+        task::spawn_blocking(move || wait_for_path_exists_blocking(path, timeout)).await?
+    }
+
+    pub async fn wait_for_matching_file(
+        root: impl Into<PathBuf>,
+        timeout: Duration,
+        predicate: impl FnMut(&Path) -> bool + Send + 'static,
+    ) -> Result<PathBuf> {
+        let root = root.into();
+        task::spawn_blocking(move || {
+            let mut predicate = predicate;
+            blocking_find_matching_file(root, timeout, &mut predicate)
+        })
+        .await?
+    }
+
+    fn wait_for_path_exists_blocking(path: PathBuf, timeout: Duration) -> Result<PathBuf> {
+        if path.exists() {
+            return Ok(path);
+        }
+
+        let watch_root = nearest_existing_ancestor(&path);
+        let (tx, rx) = mpsc::channel();
+        let mut watcher = notify::recommended_watcher(move |res| {
+            let _ = tx.send(res);
+        })?;
+        watcher.watch(&watch_root, RecursiveMode::Recursive)?;
+
+        let deadline = Instant::now() + timeout;
+        loop {
+            if path.exists() {
+                return Ok(path.clone());
+            }
+            let now = Instant::now();
+            if now >= deadline {
+                break;
+            }
+            let remaining = deadline.saturating_duration_since(now);
+            match rx.recv_timeout(remaining) {
+                Ok(Ok(_event)) => {
+                    if path.exists() {
+                        return Ok(path.clone());
+                    }
+                }
+                Ok(Err(err)) => return Err(err.into()),
+                Err(RecvTimeoutError::Timeout) => break,
+                Err(RecvTimeoutError::Disconnected) => break,
+            }
+        }
+
+        if path.exists() {
+            Ok(path)
+        } else {
+            Err(anyhow!("timed out waiting for {:?}", path))
+        }
+    }
+
+    fn blocking_find_matching_file(
+        root: PathBuf,
+        timeout: Duration,
+        predicate: &mut impl FnMut(&Path) -> bool,
+    ) -> Result<PathBuf> {
+        let root = wait_for_path_exists_blocking(root, timeout)?;
+
+        if let Some(found) = scan_for_match(&root, predicate) {
+            return Ok(found);
+        }
+
+        let (tx, rx) = mpsc::channel();
+        let mut watcher = notify::recommended_watcher(move |res| {
+            let _ = tx.send(res);
+        })?;
+        watcher.watch(&root, RecursiveMode::Recursive)?;
+
+        let deadline = Instant::now() + timeout;
+
+        while Instant::now() < deadline {
+            let remaining = deadline.saturating_duration_since(Instant::now());
+            match rx.recv_timeout(remaining) {
+                Ok(Ok(_event)) => {
+                    if let Some(found) = scan_for_match(&root, predicate) {
+                        return Ok(found);
+                    }
+                }
+                Ok(Err(err)) => return Err(err.into()),
+                Err(RecvTimeoutError::Timeout) => break,
+                Err(RecvTimeoutError::Disconnected) => break,
+            }
+        }
+
+        if let Some(found) = scan_for_match(&root, predicate) {
+            Ok(found)
+        } else {
+            Err(anyhow!("timed out waiting for matching file in {:?}", root))
+        }
+    }
+
+    fn scan_for_match(root: &Path, predicate: &mut impl FnMut(&Path) -> bool) -> Option<PathBuf> {
+        for entry in WalkDir::new(root).into_iter().filter_map(Result::ok) {
+            let path = entry.path();
+            if !entry.file_type().is_file() {
+                continue;
+            }
+            if predicate(path) {
+                return Some(path.to_path_buf());
+            }
+        }
+        None
+    }
+
+    fn nearest_existing_ancestor(path: &Path) -> PathBuf {
+        let mut current = path;
+        loop {
+            if current.exists() {
+                return current.to_path_buf();
+            }
+            match current.parent() {
+                Some(parent) => current = parent,
+                None => return PathBuf::from("."),
+            }
+        }
+    }
+}
+
 #[macro_export]
 macro_rules! skip_if_sandbox {
     () => {{

codex-rs/core/tests/common/responses.rs

@@ -1,11 +1,105 @@
+use std::sync::Arc;
+use std::sync::Mutex;
+
 use serde_json::Value;
 use wiremock::BodyPrintLimit;
+use wiremock::Match;
 use wiremock::Mock;
+use wiremock::MockBuilder;
 use wiremock::MockServer;
 use wiremock::Respond;
 use wiremock::ResponseTemplate;
 use wiremock::matchers::method;
-use wiremock::matchers::path;
+use wiremock::matchers::path_regex;
+
+#[derive(Debug, Clone)]
+pub struct ResponseMock {
+    requests: Arc<Mutex<Vec<ResponsesRequest>>>,
+}
+
+impl ResponseMock {
+    fn new() -> Self {
+        Self {
+            requests: Arc::new(Mutex::new(Vec::new())),
+        }
+    }
+
+    pub fn single_request(&self) -> ResponsesRequest {
+        let requests = self.requests.lock().unwrap();
+        if requests.len() != 1 {
+            panic!("expected 1 request, got {}", requests.len());
+        }
+        requests.first().unwrap().clone()
+    }
+
+    pub fn requests(&self) -> Vec<ResponsesRequest> {
+        self.requests.lock().unwrap().clone()
+    }
+}
+
+#[derive(Debug, Clone)]
+pub struct ResponsesRequest(wiremock::Request);
+
+impl ResponsesRequest {
+    pub fn body_json(&self) -> Value {
+        self.0.body_json().unwrap()
+    }
+
+    pub fn input(&self) -> Vec<Value> {
+        self.0.body_json::<Value>().unwrap()["input"]
+            .as_array()
+            .expect("input array not found in request")
+            .clone()
+    }
+
+    pub fn function_call_output(&self, call_id: &str) -> Value {
+        self.call_output(call_id, "function_call_output")
+    }
+
+    pub fn custom_tool_call_output(&self, call_id: &str) -> Value {
+        self.call_output(call_id, "custom_tool_call_output")
+    }
+
+    pub fn call_output(&self, call_id: &str, call_type: &str) -> Value {
+        self.input()
+            .iter()
+            .find(|item| {
+                item.get("type").unwrap() == call_type && item.get("call_id").unwrap() == call_id
+            })
+            .cloned()
+            .unwrap_or_else(|| panic!("function call output {call_id} item not found in request"))
+    }
+
+    pub fn header(&self, name: &str) -> Option<String> {
+        self.0
+            .headers
+            .get(name)
+            .and_then(|v| v.to_str().ok())
+            .map(str::to_string)
+    }
+
+    pub fn path(&self) -> String {
+        self.0.url.path().to_string()
+    }
+
+    pub fn query_param(&self, name: &str) -> Option<String> {
+        self.0
+            .url
+            .query_pairs()
+            .find(|(k, _)| k == name)
+            .map(|(_, v)| v.to_string())
+    }
+}
+
+impl Match for ResponseMock {
+    fn matches(&self, request: &wiremock::Request) -> bool {
+        self.requests
+            .lock()
+            .unwrap()
+            .push(ResponsesRequest(request.clone()));
+        true
+    }
+}
 
 /// Build an SSE stream body from a list of JSON events.
 pub fn sse(events: Vec<Value>) -> String {
@@ -34,6 +128,16 @@ pub fn ev_completed(id: &str) -> Value {
     })
 }
 
+/// Convenience: SSE event for a created response with a specific id.
+pub fn ev_response_created(id: &str) -> Value {
+    serde_json::json!({
+        "type": "response.created",
+        "response": {
+            "id": id,
+        }
+    })
+}
+
 pub fn ev_completed_with_tokens(id: &str, total_tokens: u64) -> Value {
     serde_json::json!({
         "type": "response.completed",
@@ -151,34 +255,40 @@ pub fn sse_response(body: String) -> ResponseTemplate {
         .set_body_raw(body, "text/event-stream")
 }
 
-pub async fn mount_sse_once_match<M>(server: &MockServer, matcher: M, body: String)
+fn base_mock() -> (MockBuilder, ResponseMock) {
+    let response_mock = ResponseMock::new();
+    let mock = Mock::given(method("POST"))
+        .and(path_regex(".*/responses$"))
+        .and(response_mock.clone());
+    (mock, response_mock)
+}
+
+pub async fn mount_sse_once_match<M>(server: &MockServer, matcher: M, body: String) -> ResponseMock
 where
     M: wiremock::Match + Send + Sync + 'static,
 {
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(matcher)
+    let (mock, response_mock) = base_mock();
+    mock.and(matcher)
         .respond_with(sse_response(body))
         .up_to_n_times(1)
         .mount(server)
         .await;
+    response_mock
 }
 
-pub async fn mount_sse_once(server: &MockServer, body: String) {
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(sse_response(body))
-        .expect(1)
+pub async fn mount_sse_once(server: &MockServer, body: String) -> ResponseMock {
+    let (mock, response_mock) = base_mock();
+    mock.respond_with(sse_response(body))
+        .up_to_n_times(1)
         .mount(server)
         .await;
+    response_mock
 }
 
-pub async fn mount_sse(server: &MockServer, body: String) {
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(sse_response(body))
-        .mount(server)
-        .await;
+pub async fn mount_sse(server: &MockServer, body: String) -> ResponseMock {
+    let (mock, response_mock) = base_mock();
+    mock.respond_with(sse_response(body)).mount(server).await;
+    response_mock
 }
 
 pub async fn start_mock_server() -> MockServer {
@@ -191,7 +301,7 @@ pub async fn start_mock_server() -> MockServer {
 /// Mounts a sequence of SSE response bodies and serves them in order for each
 /// POST to `/v1/responses`. Panics if more requests are received than bodies
 /// provided. Also asserts the exact number of expected calls.
-pub async fn mount_sse_sequence(server: &MockServer, bodies: Vec<String>) {
+pub async fn mount_sse_sequence(server: &MockServer, bodies: Vec<String>) -> ResponseMock {
     use std::sync::atomic::AtomicUsize;
     use std::sync::atomic::Ordering;
 
@@ -218,10 +328,11 @@ pub async fn mount_sse_sequence(server: &MockServer, bodies: Vec<String>) {
         responses: bodies,
     };
 
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(responder)
+    let (mock, response_mock) = base_mock();
+    mock.respond_with(responder)
         .expect(num_calls as u64)
         .mount(server)
         .await;
+
+    response_mock
 }

codex-rs/core/tests/common/test_codex.rs

@@ -1,4 +1,5 @@
 use std::mem::swap;
+use std::path::PathBuf;
 use std::sync::Arc;
 
 use codex_core::CodexAuth;
@@ -39,6 +40,12 @@ impl TestCodexBuilder {
         let mut config = load_default_config_for_test(&home);
         config.cwd = cwd.path().to_path_buf();
         config.model_provider = model_provider;
+        config.codex_linux_sandbox_exe = Some(PathBuf::from(
+            assert_cmd::Command::cargo_bin("codex")?
+                .get_program()
+                .to_os_string(),
+        ));
+
         let mut mutators = vec![];
         swap(&mut self.config_mutators, &mut mutators);
 

codex-rs/core/tests/responses_headers.rs

@@ -0,0 +1,102 @@
+use std::sync::Arc;
+
+use codex_app_server_protocol::AuthMode;
+use codex_core::ContentItem;
+use codex_core::ModelClient;
+use codex_core::ModelProviderInfo;
+use codex_core::Prompt;
+use codex_core::ResponseEvent;
+use codex_core::ResponseItem;
+use codex_core::WireApi;
+use codex_otel::otel_event_manager::OtelEventManager;
+use codex_protocol::ConversationId;
+use core_test_support::load_default_config_for_test;
+use core_test_support::responses;
+use futures::StreamExt;
+use tempfile::TempDir;
+use wiremock::matchers::header;
+
+#[tokio::test]
+async fn responses_stream_includes_task_type_header() {
+    core_test_support::skip_if_no_network!();
+
+    let server = responses::start_mock_server().await;
+    let response_body = responses::sse(vec![
+        responses::ev_response_created("resp-1"),
+        responses::ev_completed("resp-1"),
+    ]);
+
+    let request_recorder = responses::mount_sse_once_match(
+        &server,
+        header("Codex-Task-Type", "standard"),
+        response_body,
+    )
+    .await;
+
+    let provider = ModelProviderInfo {
+        name: "mock".into(),
+        base_url: Some(format!("{}/v1", server.uri())),
+        env_key: None,
+        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(5_000),
+        requires_openai_auth: false,
+    };
+
+    let codex_home = TempDir::new().expect("failed to create TempDir");
+    let mut config = load_default_config_for_test(&codex_home);
+    config.model_provider_id = provider.name.clone();
+    config.model_provider = provider.clone();
+    let effort = config.model_reasoning_effort;
+    let summary = config.model_reasoning_summary;
+    let config = Arc::new(config);
+
+    let conversation_id = ConversationId::new();
+
+    let otel_event_manager = OtelEventManager::new(
+        conversation_id,
+        config.model.as_str(),
+        config.model_family.slug.as_str(),
+        None,
+        Some(AuthMode::ChatGPT),
+        false,
+        "test".to_string(),
+    );
+
+    let client = ModelClient::new(
+        Arc::clone(&config),
+        None,
+        otel_event_manager,
+        provider,
+        effort,
+        summary,
+        conversation_id,
+    );
+
+    let mut prompt = Prompt::default();
+    prompt.input = vec![ResponseItem::Message {
+        id: None,
+        role: "user".into(),
+        content: vec![ContentItem::InputText {
+            text: "hello".into(),
+        }],
+    }];
+
+    let mut stream = client.stream(&prompt).await.expect("stream failed");
+    while let Some(event) = stream.next().await {
+        if matches!(event, Ok(ResponseEvent::Completed { .. })) {
+            break;
+        }
+    }
+
+    let request = request_recorder.single_request();
+    assert_eq!(
+        request.header("Codex-Task-Type").as_deref(),
+        Some("standard")
+    );
+}

codex-rs/core/tests/suite/cli_stream.rs

@@ -1,12 +1,11 @@
 use assert_cmd::Command as AssertCommand;
 use codex_core::RolloutRecorder;
 use codex_core::protocol::GitInfo;
+use core_test_support::fs_wait;
 use core_test_support::skip_if_no_network;
 use std::time::Duration;
-use std::time::Instant;
 use tempfile::TempDir;
 use uuid::Uuid;
-use walkdir::WalkDir;
 use wiremock::Mock;
 use wiremock::MockServer;
 use wiremock::ResponseTemplate;
@@ -106,16 +105,12 @@ async fn exec_cli_applies_experimental_instructions_file() {
         "data: {\"type\":\"response.created\",\"response\":{}}\n\n",
         "data: {\"type\":\"response.completed\",\"response\":{\"id\":\"r1\"}}\n\n"
     );
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(
-            ResponseTemplate::new(200)
-                .insert_header("content-type", "text/event-stream")
-                .set_body_raw(sse, "text/event-stream"),
-        )
-        .expect(1)
-        .mount(&server)
-        .await;
+    let resp_mock = core_test_support::responses::mount_sse_once_match(
+        &server,
+        path("/v1/responses"),
+        sse.to_string(),
+    )
+    .await;
 
     // Create a temporary instructions file with a unique marker we can assert
     // appears in the outbound request payload.
@@ -164,8 +159,8 @@ async fn exec_cli_applies_experimental_instructions_file() {
 
     // Inspect the captured request and verify our custom base instructions were
     // included in the `instructions` field.
-    let request = &server.received_requests().await.unwrap()[0];
-    let body = request.body_json::<serde_json::Value>().unwrap();
+    let request = resp_mock.single_request();
+    let body = request.body_json();
     let instructions = body
         .get("instructions")
         .and_then(|v| v.as_str())
@@ -215,12 +210,12 @@ async fn responses_api_stream_cli() {
 
 /// End-to-end: create a session (writes rollout), verify the file, then resume and confirm append.
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn integration_creates_and_checks_session_file() {
+async fn integration_creates_and_checks_session_file() -> anyhow::Result<()> {
     // Honor sandbox network restrictions for CI parity with the other tests.
-    skip_if_no_network!();
+    skip_if_no_network!(Ok(()));
 
     // 1. Temp home so we read/write isolated session files.
-    let home = TempDir::new().unwrap();
+    let home = TempDir::new()?;
 
     // 2. Unique marker we'll look for in the session log.
     let marker = format!("integration-test-{}", Uuid::new_v4());
@@ -258,63 +253,20 @@ async fn integration_creates_and_checks_session_file() {
 
     // Wait for sessions dir to appear.
     let sessions_dir = home.path().join("sessions");
-    let dir_deadline = Instant::now() + Duration::from_secs(5);
-    while !sessions_dir.exists() && Instant::now() < dir_deadline {
-        std::thread::sleep(Duration::from_millis(50));
-    }
-    assert!(sessions_dir.exists(), "sessions directory never appeared");
+    fs_wait::wait_for_path_exists(&sessions_dir, Duration::from_secs(5)).await?;
 
     // Find the session file that contains `marker`.
-    let deadline = Instant::now() + Duration::from_secs(10);
-    let mut matching_path: Option<std::path::PathBuf> = None;
-    while Instant::now() < deadline && matching_path.is_none() {
-        for entry in WalkDir::new(&sessions_dir) {
-            let entry = match entry {
-                Ok(e) => e,
-                Err(_) => continue,
-            };
-            if !entry.file_type().is_file() {
-                continue;
-            }
-            if !entry.file_name().to_string_lossy().ends_with(".jsonl") {
-                continue;
-            }
-            let path = entry.path();
-            let Ok(content) = std::fs::read_to_string(path) else {
-                continue;
-            };
-            let mut lines = content.lines();
-            if lines.next().is_none() {
-                continue;
-            }
-            for line in lines {
-                if line.trim().is_empty() {
-                    continue;
-                }
-                let item: serde_json::Value = match serde_json::from_str(line) {
-                    Ok(v) => v,
-                    Err(_) => continue,
-                };
-                if item.get("type").and_then(|t| t.as_str()) == Some("response_item")
-                    && let Some(payload) = item.get("payload")
-                    && payload.get("type").and_then(|t| t.as_str()) == Some("message")
-                    && let Some(c) = payload.get("content")
-                    && c.to_string().contains(&marker)
-                {
-                    matching_path = Some(path.to_path_buf());
-                    break;
-                }
-            }
+    let marker_clone = marker.clone();
+    let path = fs_wait::wait_for_matching_file(&sessions_dir, Duration::from_secs(10), move |p| {
+        if p.extension().and_then(|ext| ext.to_str()) != Some("jsonl") {
+            return false;
         }
-        if matching_path.is_none() {
-            std::thread::sleep(Duration::from_millis(50));
-        }
-    }
-
-    let path = match matching_path {
-        Some(p) => p,
-        None => panic!("No session file containing the marker was found"),
-    };
+        let Ok(content) = std::fs::read_to_string(p) else {
+            return false;
+        };
+        content.contains(&marker_clone)
+    })
+    .await?;
 
     // Basic sanity checks on location and metadata.
     let rel = match path.strip_prefix(&sessions_dir) {
@@ -422,42 +374,25 @@ async fn integration_creates_and_checks_session_file() {
     assert!(output2.status.success(), "resume codex-cli run failed");
 
     // Find the new session file containing the resumed marker.
-    let deadline = Instant::now() + Duration::from_secs(10);
-    let mut resumed_path: Option<std::path::PathBuf> = None;
-    while Instant::now() < deadline && resumed_path.is_none() {
-        for entry in WalkDir::new(&sessions_dir) {
-            let entry = match entry {
-                Ok(e) => e,
-                Err(_) => continue,
-            };
-            if !entry.file_type().is_file() {
-                continue;
+    let marker2_clone = marker2.clone();
+    let resumed_path =
+        fs_wait::wait_for_matching_file(&sessions_dir, Duration::from_secs(10), move |p| {
+            if p.extension().and_then(|ext| ext.to_str()) != Some("jsonl") {
+                return false;
             }
-            if !entry.file_name().to_string_lossy().ends_with(".jsonl") {
-                continue;
-            }
-            let p = entry.path();
-            let Ok(c) = std::fs::read_to_string(p) else {
-                continue;
-            };
-            if c.contains(&marker2) {
-                resumed_path = Some(p.to_path_buf());
-                break;
-            }
-        }
-        if resumed_path.is_none() {
-            std::thread::sleep(Duration::from_millis(50));
-        }
-    }
+            std::fs::read_to_string(p)
+                .map(|content| content.contains(&marker2_clone))
+                .unwrap_or(false)
+        })
+        .await?;
 
-    let resumed_path = resumed_path.expect("No resumed session file found containing the marker2");
     // Resume should write to the existing log file.
     assert_eq!(
         resumed_path, path,
         "resume should create a new session file"
     );
 
-    let resumed_content = std::fs::read_to_string(&resumed_path).unwrap();
+    let resumed_content = std::fs::read_to_string(&resumed_path)?;
     assert!(
         resumed_content.contains(&marker),
         "resumed file missing original marker"
@@ -466,6 +401,7 @@ async fn integration_creates_and_checks_session_file() {
         resumed_content.contains(&marker2),
         "resumed file missing resumed marker"
     );
+    Ok(())
 }
 
 /// Integration test to verify git info is collected and recorded in session files.

codex-rs/core/tests/suite/client.rs

@@ -223,15 +223,9 @@ async fn resume_includes_initial_messages_and_sends_prior_items() {
 
     // Mock server that will receive the resumed request
     let server = MockServer::start().await;
-    let first = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse_completed("resp1"), "text/event-stream");
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(first)
-        .expect(1)
-        .mount(&server)
-        .await;
+    let resp_mock =
+        responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
+            .await;
 
     // Configure Codex to resume from our file
     let model_provider = ModelProviderInfo {
@@ -277,8 +271,8 @@ async fn resume_includes_initial_messages_and_sends_prior_items() {
         .unwrap();
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let request = &server.received_requests().await.unwrap()[0];
-    let request_body = request.body_json::<serde_json::Value>().unwrap();
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
     let expected_input = json!([
         {
             "type": "message",
@@ -372,18 +366,9 @@ async fn includes_base_instructions_override_in_request() {
     skip_if_no_network!();
     // Mock server
     let server = MockServer::start().await;
-
-    // First request – must NOT include `previous_response_id`.
-    let first = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse_completed("resp1"), "text/event-stream");
-
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(first)
-        .expect(1)
-        .mount(&server)
-        .await;
+    let resp_mock =
+        responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
+            .await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -414,8 +399,8 @@ async fn includes_base_instructions_override_in_request() {
 
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let request = &server.received_requests().await.unwrap()[0];
-    let request_body = request.body_json::<serde_json::Value>().unwrap();
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
 
     assert!(
         request_body["instructions"]
@@ -570,16 +555,9 @@ async fn includes_user_instructions_message_in_request() {
     skip_if_no_network!();
     let server = MockServer::start().await;
 
-    let first = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse_completed("resp1"), "text/event-stream");
-
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(first)
-        .expect(1)
-        .mount(&server)
-        .await;
+    let resp_mock =
+        responses::mount_sse_once_match(&server, path("/v1/responses"), sse_completed("resp1"))
+            .await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -610,8 +588,8 @@ async fn includes_user_instructions_message_in_request() {
 
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let request = &server.received_requests().await.unwrap()[0];
-    let request_body = request.body_json::<serde_json::Value>().unwrap();
+    let request = resp_mock.single_request();
+    let request_body = request.body_json();
 
     assert!(
         !request_body["instructions"]

codex-rs/core/tests/suite/compact.rs

@@ -13,12 +13,6 @@ use core_test_support::load_default_config_for_test;
 use core_test_support::skip_if_no_network;
 use core_test_support::wait_for_event;
 use tempfile::TempDir;
-use wiremock::Mock;
-use wiremock::Request;
-use wiremock::Respond;
-use wiremock::ResponseTemplate;
-use wiremock::matchers::method;
-use wiremock::matchers::path;
 
 use codex_core::codex::compact::SUMMARIZATION_PROMPT;
 use core_test_support::responses::ev_assistant_message;
@@ -26,14 +20,11 @@ use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_completed_with_tokens;
 use core_test_support::responses::ev_function_call;
 use core_test_support::responses::mount_sse_once_match;
+use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
-use core_test_support::responses::sse_response;
+use core_test_support::responses::sse_failed;
 use core_test_support::responses::start_mock_server;
 use pretty_assertions::assert_eq;
-use std::sync::Arc;
-use std::sync::Mutex;
-use std::sync::atomic::AtomicUsize;
-use std::sync::atomic::Ordering;
 // --- Test helpers -----------------------------------------------------------
 
 pub(super) const FIRST_REPLY: &str = "FIRST_REPLY";
@@ -48,6 +39,8 @@ const SECOND_LARGE_REPLY: &str = "SECOND_LARGE_REPLY";
 const FIRST_AUTO_SUMMARY: &str = "FIRST_AUTO_SUMMARY";
 const SECOND_AUTO_SUMMARY: &str = "SECOND_AUTO_SUMMARY";
 const FINAL_REPLY: &str = "FINAL_REPLY";
+const CONTEXT_LIMIT_MESSAGE: &str =
+    "Your input exceeds the context window of this model. Please adjust your input and try again.";
 const DUMMY_FUNCTION_NAME: &str = "unsupported_tool";
 const DUMMY_CALL_ID: &str = "call-multi-auto";
 
@@ -295,12 +288,7 @@ async fn auto_compact_runs_after_token_limit_hit() {
             && !body.contains(SECOND_AUTO_MSG)
             && !body.contains("You have exceeded the maximum number of tokens")
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(first_matcher)
-        .respond_with(sse_response(sse1))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, first_matcher, sse1).await;
 
     let second_matcher = |req: &wiremock::Request| {
         let body = std::str::from_utf8(&req.body).unwrap_or("");
@@ -308,23 +296,13 @@ async fn auto_compact_runs_after_token_limit_hit() {
             && body.contains(FIRST_AUTO_MSG)
             && !body.contains("You have exceeded the maximum number of tokens")
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(second_matcher)
-        .respond_with(sse_response(sse2))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, second_matcher, sse2).await;
 
     let third_matcher = |req: &wiremock::Request| {
         let body = std::str::from_utf8(&req.body).unwrap_or("");
         body.contains("You have exceeded the maximum number of tokens")
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(third_matcher)
-        .respond_with(sse_response(sse3))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, third_matcher, sse3).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -455,12 +433,7 @@ async fn auto_compact_persists_rollout_entries() {
             && !body.contains(SECOND_AUTO_MSG)
             && !body.contains("You have exceeded the maximum number of tokens")
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(first_matcher)
-        .respond_with(sse_response(sse1))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, first_matcher, sse1).await;
 
     let second_matcher = |req: &wiremock::Request| {
         let body = std::str::from_utf8(&req.body).unwrap_or("");
@@ -468,23 +441,13 @@ async fn auto_compact_persists_rollout_entries() {
             && body.contains(FIRST_AUTO_MSG)
             && !body.contains("You have exceeded the maximum number of tokens")
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(second_matcher)
-        .respond_with(sse_response(sse2))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, second_matcher, sse2).await;
 
     let third_matcher = |req: &wiremock::Request| {
         let body = std::str::from_utf8(&req.body).unwrap_or("");
         body.contains("You have exceeded the maximum number of tokens")
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(third_matcher)
-        .respond_with(sse_response(sse3))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, third_matcher, sse3).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -582,35 +545,20 @@ async fn auto_compact_stops_after_failed_attempt() {
         body.contains(FIRST_AUTO_MSG)
             && !body.contains("You have exceeded the maximum number of tokens")
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(first_matcher)
-        .respond_with(sse_response(sse1.clone()))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, first_matcher, sse1.clone()).await;
 
     let second_matcher = |req: &wiremock::Request| {
         let body = std::str::from_utf8(&req.body).unwrap_or("");
         body.contains("You have exceeded the maximum number of tokens")
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(second_matcher)
-        .respond_with(sse_response(sse2.clone()))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, second_matcher, sse2.clone()).await;
 
     let third_matcher = |req: &wiremock::Request| {
         let body = std::str::from_utf8(&req.body).unwrap_or("");
         !body.contains("You have exceeded the maximum number of tokens")
             && body.contains(SUMMARY_TEXT)
     };
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(third_matcher)
-        .respond_with(sse_response(sse3.clone()))
-        .mount(&server)
-        .await;
+    mount_sse_once_match(&server, third_matcher, sse3.clone()).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -677,6 +625,130 @@ async fn auto_compact_stops_after_failed_attempt() {
     );
 }
 
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn manual_compact_retries_after_context_window_error() {
+    skip_if_no_network!();
+
+    let server = start_mock_server().await;
+
+    let user_turn = sse(vec![
+        ev_assistant_message("m1", FIRST_REPLY),
+        ev_completed("r1"),
+    ]);
+    let compact_failed = sse_failed(
+        "resp-fail",
+        "context_length_exceeded",
+        CONTEXT_LIMIT_MESSAGE,
+    );
+    let compact_succeeds = sse(vec![
+        ev_assistant_message("m2", SUMMARY_TEXT),
+        ev_completed("r2"),
+    ]);
+
+    let request_log = mount_sse_sequence(
+        &server,
+        vec![
+            user_turn.clone(),
+            compact_failed.clone(),
+            compact_succeeds.clone(),
+        ],
+    )
+    .await;
+
+    let model_provider = ModelProviderInfo {
+        base_url: Some(format!("{}/v1", server.uri())),
+        ..built_in_model_providers()["openai"].clone()
+    };
+
+    let home = TempDir::new().unwrap();
+    let mut config = load_default_config_for_test(&home);
+    config.model_provider = model_provider;
+    config.model_auto_compact_token_limit = Some(200_000);
+    let codex = ConversationManager::with_auth(CodexAuth::from_api_key("dummy"))
+        .new_conversation(config)
+        .await
+        .unwrap()
+        .conversation;
+
+    codex
+        .submit(Op::UserInput {
+            items: vec![InputItem::Text {
+                text: "first turn".into(),
+            }],
+        })
+        .await
+        .unwrap();
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    codex.submit(Op::Compact).await.unwrap();
+
+    let EventMsg::BackgroundEvent(event) =
+        wait_for_event(&codex, |ev| matches!(ev, EventMsg::BackgroundEvent(_))).await
+    else {
+        panic!("expected background event after compact retry");
+    };
+    assert!(
+        event.message.contains("Trimmed 1 older conversation item"),
+        "background event should mention trimmed item count: {}",
+        event.message
+    );
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let requests = request_log.requests();
+    assert_eq!(
+        requests.len(),
+        3,
+        "expected user turn and two compact attempts"
+    );
+
+    let compact_attempt = requests[1].body_json();
+    let retry_attempt = requests[2].body_json();
+
+    let compact_input = compact_attempt["input"]
+        .as_array()
+        .unwrap_or_else(|| panic!("compact attempt missing input array: {compact_attempt}"));
+    let retry_input = retry_attempt["input"]
+        .as_array()
+        .unwrap_or_else(|| panic!("retry attempt missing input array: {retry_attempt}"));
+    assert_eq!(
+        compact_input
+            .last()
+            .and_then(|item| item.get("content"))
+            .and_then(|v| v.as_array())
+            .and_then(|items| items.first())
+            .and_then(|entry| entry.get("text"))
+            .and_then(|text| text.as_str()),
+        Some(SUMMARIZATION_PROMPT),
+        "compact attempt should include summarization prompt"
+    );
+    assert_eq!(
+        retry_input
+            .last()
+            .and_then(|item| item.get("content"))
+            .and_then(|v| v.as_array())
+            .and_then(|items| items.first())
+            .and_then(|entry| entry.get("text"))
+            .and_then(|text| text.as_str()),
+        Some(SUMMARIZATION_PROMPT),
+        "retry attempt should include summarization prompt"
+    );
+    assert_eq!(
+        retry_input.len(),
+        compact_input.len().saturating_sub(1),
+        "retry should drop exactly one history item (before {} vs after {})",
+        compact_input.len(),
+        retry_input.len()
+    );
+    if let (Some(first_before), Some(first_after)) = (compact_input.first(), retry_input.first()) {
+        assert_ne!(
+            first_before, first_after,
+            "retry should drop the oldest conversation item"
+        );
+    } else {
+        panic!("expected non-empty compact inputs");
+    }
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn auto_compact_allows_multiple_attempts_when_interleaved_with_other_turn_events() {
     skip_if_no_network!();
@@ -708,49 +780,7 @@ async fn auto_compact_allows_multiple_attempts_when_interleaved_with_other_turn_
         ev_completed_with_tokens("r6", 120),
     ]);
 
-    #[derive(Clone)]
-    struct SeqResponder {
-        bodies: Arc<Vec<String>>,
-        calls: Arc<AtomicUsize>,
-        requests: Arc<Mutex<Vec<Vec<u8>>>>,
-    }
-
-    impl SeqResponder {
-        fn new(bodies: Vec<String>) -> Self {
-            Self {
-                bodies: Arc::new(bodies),
-                calls: Arc::new(AtomicUsize::new(0)),
-                requests: Arc::new(Mutex::new(Vec::new())),
-            }
-        }
-
-        fn recorded_requests(&self) -> Vec<Vec<u8>> {
-            self.requests.lock().unwrap().clone()
-        }
-    }
-
-    impl Respond for SeqResponder {
-        fn respond(&self, req: &Request) -> ResponseTemplate {
-            let idx = self.calls.fetch_add(1, Ordering::SeqCst);
-            self.requests.lock().unwrap().push(req.body.clone());
-            let body = self
-                .bodies
-                .get(idx)
-                .unwrap_or_else(|| panic!("unexpected request index {idx}"))
-                .clone();
-            ResponseTemplate::new(200)
-                .insert_header("content-type", "text/event-stream")
-                .set_body_raw(body, "text/event-stream")
-        }
-    }
-
-    let responder = SeqResponder::new(vec![sse1, sse2, sse3, sse4, sse5, sse6]);
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(responder.clone())
-        .expect(6)
-        .mount(&server)
-        .await;
+    mount_sse_sequence(&server, vec![sse1, sse2, sse3, sse4, sse5, sse6]).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -801,10 +831,12 @@ async fn auto_compact_allows_multiple_attempts_when_interleaved_with_other_turn_
         "auto compact should not emit task lifecycle events"
     );
 
-    let request_bodies: Vec<String> = responder
-        .recorded_requests()
+    let request_bodies: Vec<String> = server
+        .received_requests()
+        .await
+        .unwrap()
         .into_iter()
-        .map(|body| String::from_utf8(body).unwrap_or_default())
+        .map(|request| String::from_utf8(request.body).unwrap_or_default())
         .collect();
     assert_eq!(
         request_bodies.len(),

codex-rs/core/tests/suite/grep_files.rs

@@ -0,0 +1,237 @@
+#![cfg(not(target_os = "windows"))]
+
+use anyhow::Result;
+use codex_core::model_family::find_family_for_model;
+use codex_core::protocol::AskForApproval;
+use codex_core::protocol::EventMsg;
+use codex_core::protocol::InputItem;
+use codex_core::protocol::Op;
+use codex_core::protocol::SandboxPolicy;
+use codex_protocol::config_types::ReasoningSummary;
+use core_test_support::responses;
+use core_test_support::responses::ev_assistant_message;
+use core_test_support::responses::ev_completed;
+use core_test_support::responses::ev_function_call;
+use core_test_support::responses::ev_response_created;
+use core_test_support::responses::sse;
+use core_test_support::responses::start_mock_server;
+use core_test_support::skip_if_no_network;
+use core_test_support::test_codex::TestCodex;
+use core_test_support::test_codex::test_codex;
+use core_test_support::wait_for_event;
+use serde_json::Value;
+use std::collections::HashSet;
+use std::path::Path;
+use std::process::Command as StdCommand;
+use wiremock::matchers::any;
+
+const MODEL_WITH_TOOL: &str = "test-gpt-5-codex";
+
+fn ripgrep_available() -> bool {
+    StdCommand::new("rg")
+        .arg("--version")
+        .output()
+        .map(|output| output.status.success())
+        .unwrap_or(false)
+}
+
+macro_rules! skip_if_ripgrep_missing {
+    ($ret:expr $(,)?) => {{
+        if !ripgrep_available() {
+            eprintln!("rg not available in PATH; skipping test");
+            return $ret;
+        }
+    }};
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn grep_files_tool_collects_matches() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+    skip_if_ripgrep_missing!(Ok(()));
+
+    let server = start_mock_server().await;
+    let test = build_test_codex(&server).await?;
+
+    let search_dir = test.cwd.path().join("src");
+    std::fs::create_dir_all(&search_dir)?;
+    let alpha = search_dir.join("alpha.rs");
+    let beta = search_dir.join("beta.rs");
+    let gamma = search_dir.join("gamma.txt");
+    std::fs::write(&alpha, "alpha needle\n")?;
+    std::fs::write(&beta, "beta needle\n")?;
+    std::fs::write(&gamma, "needle in text but excluded\n")?;
+
+    let call_id = "grep-files-collect";
+    let arguments = serde_json::json!({
+        "pattern": "needle",
+        "path": search_dir.to_string_lossy(),
+        "include": "*.rs",
+    })
+    .to_string();
+
+    mount_tool_sequence(&server, call_id, &arguments, "grep_files").await;
+    submit_turn(&test, "please find uses of needle").await?;
+
+    let bodies = recorded_bodies(&server).await?;
+    let tool_output = find_tool_output(&bodies, call_id).expect("tool output present");
+    let payload = tool_output.get("output").expect("output field present");
+    let (content_opt, success_opt) = extract_content_and_success(payload);
+    let content = content_opt.expect("content present");
+    let success = success_opt.unwrap_or(true);
+    assert!(success, "expected success for matches, got {payload:?}");
+
+    let entries = collect_file_names(content);
+    assert_eq!(entries.len(), 2, "content: {content}");
+    assert!(
+        entries.contains("alpha.rs"),
+        "missing alpha.rs in {entries:?}"
+    );
+    assert!(
+        entries.contains("beta.rs"),
+        "missing beta.rs in {entries:?}"
+    );
+    assert!(
+        !entries.contains("gamma.txt"),
+        "txt file should be filtered out: {entries:?}"
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn grep_files_tool_reports_empty_results() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+    skip_if_ripgrep_missing!(Ok(()));
+
+    let server = start_mock_server().await;
+    let test = build_test_codex(&server).await?;
+
+    let search_dir = test.cwd.path().join("logs");
+    std::fs::create_dir_all(&search_dir)?;
+    std::fs::write(search_dir.join("output.txt"), "no hits here")?;
+
+    let call_id = "grep-files-empty";
+    let arguments = serde_json::json!({
+        "pattern": "needle",
+        "path": search_dir.to_string_lossy(),
+        "limit": 5,
+    })
+    .to_string();
+
+    mount_tool_sequence(&server, call_id, &arguments, "grep_files").await;
+    submit_turn(&test, "search again").await?;
+
+    let bodies = recorded_bodies(&server).await?;
+    let tool_output = find_tool_output(&bodies, call_id).expect("tool output present");
+    let payload = tool_output.get("output").expect("output field present");
+    let (content_opt, success_opt) = extract_content_and_success(payload);
+    let content = content_opt.expect("content present");
+    if let Some(success) = success_opt {
+        assert!(!success, "expected success=false payload: {payload:?}");
+    }
+    assert_eq!(content, "No matches found.");
+
+    Ok(())
+}
+
+#[allow(clippy::expect_used)]
+async fn build_test_codex(server: &wiremock::MockServer) -> Result<TestCodex> {
+    let mut builder = test_codex().with_config(|config| {
+        config.model = MODEL_WITH_TOOL.to_string();
+        config.model_family =
+            find_family_for_model(MODEL_WITH_TOOL).expect("model family for test model");
+    });
+    builder.build(server).await
+}
+
+async fn submit_turn(test: &TestCodex, prompt: &str) -> Result<()> {
+    let session_model = test.session_configured.model.clone();
+
+    test.codex
+        .submit(Op::UserTurn {
+            items: vec![InputItem::Text {
+                text: prompt.into(),
+            }],
+            final_output_json_schema: None,
+            cwd: test.cwd.path().to_path_buf(),
+            approval_policy: AskForApproval::Never,
+            sandbox_policy: SandboxPolicy::DangerFullAccess,
+            model: session_model,
+            effort: None,
+            summary: ReasoningSummary::Auto,
+        })
+        .await?;
+
+    wait_for_event(&test.codex, |event| {
+        matches!(event, EventMsg::TaskComplete(_))
+    })
+    .await;
+    Ok(())
+}
+
+async fn mount_tool_sequence(
+    server: &wiremock::MockServer,
+    call_id: &str,
+    arguments: &str,
+    tool_name: &str,
+) {
+    let first_response = sse(vec![
+        ev_response_created("resp-1"),
+        ev_function_call(call_id, tool_name, arguments),
+        ev_completed("resp-1"),
+    ]);
+    responses::mount_sse_once_match(server, any(), first_response).await;
+
+    let second_response = sse(vec![
+        ev_assistant_message("msg-1", "done"),
+        ev_completed("resp-2"),
+    ]);
+    responses::mount_sse_once_match(server, any(), second_response).await;
+}
+
+#[allow(clippy::expect_used)]
+async fn recorded_bodies(server: &wiremock::MockServer) -> Result<Vec<Value>> {
+    let requests = server.received_requests().await.expect("requests recorded");
+    Ok(requests
+        .iter()
+        .map(|req| req.body_json::<Value>().expect("request json"))
+        .collect())
+}
+
+fn find_tool_output<'a>(requests: &'a [Value], call_id: &str) -> Option<&'a Value> {
+    requests.iter().find_map(|body| {
+        body.get("input")
+            .and_then(Value::as_array)
+            .and_then(|items| {
+                items.iter().find(|item| {
+                    item.get("type").and_then(Value::as_str) == Some("function_call_output")
+                        && item.get("call_id").and_then(Value::as_str) == Some(call_id)
+                })
+            })
+    })
+}
+
+fn collect_file_names(content: &str) -> HashSet<String> {
+    content
+        .lines()
+        .filter_map(|line| {
+            if line.trim().is_empty() {
+                return None;
+            }
+            Path::new(line)
+                .file_name()
+                .map(|name| name.to_string_lossy().into_owned())
+        })
+        .collect()
+}
+
+fn extract_content_and_success(value: &Value) -> (Option<&str>, Option<bool>) {
+    match value {
+        Value::String(text) => (Some(text.as_str()), None),
+        Value::Object(obj) => (
+            obj.get("content").and_then(Value::as_str),
+            obj.get("success").and_then(Value::as_bool),
+        ),
+        _ => (None, None),
+    }
+}

codex-rs/core/tests/suite/list_dir.rs

@@ -0,0 +1,460 @@
+#![cfg(not(target_os = "windows"))]
+
+use codex_core::protocol::AskForApproval;
+use codex_core::protocol::EventMsg;
+use codex_core::protocol::InputItem;
+use codex_core::protocol::Op;
+use codex_core::protocol::SandboxPolicy;
+use codex_protocol::config_types::ReasoningSummary;
+use core_test_support::responses;
+use core_test_support::responses::ev_assistant_message;
+use core_test_support::responses::ev_completed;
+use core_test_support::responses::ev_function_call;
+use core_test_support::responses::ev_response_created;
+use core_test_support::responses::sse;
+use core_test_support::responses::start_mock_server;
+use core_test_support::skip_if_no_network;
+use core_test_support::test_codex::TestCodex;
+use core_test_support::test_codex::test_codex;
+use core_test_support::wait_for_event;
+use pretty_assertions::assert_eq;
+use serde_json::Value;
+use wiremock::matchers::any;
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[ignore = "disabled until we enable list_dir tool"]
+async fn list_dir_tool_returns_entries() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+
+    let TestCodex {
+        codex,
+        cwd,
+        session_configured,
+        ..
+    } = test_codex().build(&server).await?;
+
+    let dir_path = cwd.path().join("sample_dir");
+    std::fs::create_dir(&dir_path)?;
+    std::fs::write(dir_path.join("alpha.txt"), "first file")?;
+    std::fs::create_dir(dir_path.join("nested"))?;
+    let dir_path = dir_path.to_string_lossy().to_string();
+
+    let call_id = "list-dir-call";
+    let arguments = serde_json::json!({
+        "dir_path": dir_path,
+        "offset": 1,
+        "limit": 2,
+    })
+    .to_string();
+
+    let first_response = sse(vec![
+        ev_response_created("resp-1"),
+        ev_function_call(call_id, "list_dir", &arguments),
+        ev_completed("resp-1"),
+    ]);
+    responses::mount_sse_once_match(&server, any(), first_response).await;
+
+    let second_response = sse(vec![
+        ev_assistant_message("msg-1", "done"),
+        ev_completed("resp-2"),
+    ]);
+    responses::mount_sse_once_match(&server, any(), second_response).await;
+
+    let session_model = session_configured.model.clone();
+
+    codex
+        .submit(Op::UserTurn {
+            items: vec![InputItem::Text {
+                text: "list directory contents".into(),
+            }],
+            final_output_json_schema: None,
+            cwd: cwd.path().to_path_buf(),
+            approval_policy: AskForApproval::Never,
+            sandbox_policy: SandboxPolicy::DangerFullAccess,
+            model: session_model,
+            effort: None,
+            summary: ReasoningSummary::Auto,
+        })
+        .await?;
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let requests = server.received_requests().await.expect("recorded requests");
+    let request_bodies = requests
+        .iter()
+        .map(|req| req.body_json::<Value>().unwrap())
+        .collect::<Vec<_>>();
+    assert!(
+        !request_bodies.is_empty(),
+        "expected at least one request body"
+    );
+
+    let tool_output_item = request_bodies
+        .iter()
+        .find_map(|body| {
+            body.get("input")
+                .and_then(Value::as_array)
+                .and_then(|items| {
+                    items.iter().find(|item| {
+                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
+                    })
+                })
+        })
+        .unwrap_or_else(|| {
+            panic!("function_call_output item not found in requests: {request_bodies:#?}")
+        });
+
+    assert_eq!(
+        tool_output_item.get("call_id").and_then(Value::as_str),
+        Some(call_id)
+    );
+
+    let output_text = tool_output_item
+        .get("output")
+        .and_then(|value| match value {
+            Value::String(text) => Some(text.as_str()),
+            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
+            _ => None,
+        })
+        .expect("output text present");
+    assert_eq!(output_text, "E1: [file] alpha.txt\nE2: [dir] nested");
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[ignore = "disabled until we enable list_dir tool"]
+async fn list_dir_tool_depth_one_omits_children() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+
+    let TestCodex {
+        codex,
+        cwd,
+        session_configured,
+        ..
+    } = test_codex().build(&server).await?;
+
+    let dir_path = cwd.path().join("depth_one");
+    std::fs::create_dir(&dir_path)?;
+    std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
+    std::fs::create_dir(dir_path.join("nested"))?;
+    std::fs::write(dir_path.join("nested").join("beta.txt"), "beta")?;
+    let dir_path = dir_path.to_string_lossy().to_string();
+
+    let call_id = "list-dir-depth1";
+    let arguments = serde_json::json!({
+        "dir_path": dir_path,
+        "offset": 1,
+        "limit": 10,
+        "depth": 1,
+    })
+    .to_string();
+
+    let first_response = sse(vec![
+        ev_response_created("resp-1"),
+        ev_function_call(call_id, "list_dir", &arguments),
+        ev_completed("resp-1"),
+    ]);
+    responses::mount_sse_once_match(&server, any(), first_response).await;
+
+    let second_response = sse(vec![
+        ev_assistant_message("msg-1", "done"),
+        ev_completed("resp-2"),
+    ]);
+    responses::mount_sse_once_match(&server, any(), second_response).await;
+
+    let session_model = session_configured.model.clone();
+
+    codex
+        .submit(Op::UserTurn {
+            items: vec![InputItem::Text {
+                text: "list directory contents depth one".into(),
+            }],
+            final_output_json_schema: None,
+            cwd: cwd.path().to_path_buf(),
+            approval_policy: AskForApproval::Never,
+            sandbox_policy: SandboxPolicy::DangerFullAccess,
+            model: session_model,
+            effort: None,
+            summary: ReasoningSummary::Auto,
+        })
+        .await?;
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let requests = server.received_requests().await.expect("recorded requests");
+    let request_bodies = requests
+        .iter()
+        .map(|req| req.body_json::<Value>().unwrap())
+        .collect::<Vec<_>>();
+    assert!(
+        !request_bodies.is_empty(),
+        "expected at least one request body"
+    );
+
+    let tool_output_item = request_bodies
+        .iter()
+        .find_map(|body| {
+            body.get("input")
+                .and_then(Value::as_array)
+                .and_then(|items| {
+                    items.iter().find(|item| {
+                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
+                    })
+                })
+        })
+        .unwrap_or_else(|| {
+            panic!("function_call_output item not found in requests: {request_bodies:#?}")
+        });
+
+    assert_eq!(
+        tool_output_item.get("call_id").and_then(Value::as_str),
+        Some(call_id)
+    );
+
+    let output_text = tool_output_item
+        .get("output")
+        .and_then(|value| match value {
+            Value::String(text) => Some(text.as_str()),
+            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
+            _ => None,
+        })
+        .expect("output text present");
+    assert_eq!(output_text, "E1: [file] alpha.txt\nE2: [dir] nested");
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[ignore = "disabled until we enable list_dir tool"]
+async fn list_dir_tool_depth_two_includes_children_only() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+
+    let TestCodex {
+        codex,
+        cwd,
+        session_configured,
+        ..
+    } = test_codex().build(&server).await?;
+
+    let dir_path = cwd.path().join("depth_two");
+    std::fs::create_dir(&dir_path)?;
+    std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
+    let nested = dir_path.join("nested");
+    std::fs::create_dir(&nested)?;
+    std::fs::write(nested.join("beta.txt"), "beta")?;
+    let deeper = nested.join("grand");
+    std::fs::create_dir(&deeper)?;
+    std::fs::write(deeper.join("gamma.txt"), "gamma")?;
+    let dir_path_string = dir_path.to_string_lossy().to_string();
+
+    let call_id = "list-dir-depth2";
+    let arguments = serde_json::json!({
+        "dir_path": dir_path_string,
+        "offset": 1,
+        "limit": 10,
+        "depth": 2,
+    })
+    .to_string();
+
+    let first_response = sse(vec![
+        serde_json::json!({
+            "type": "response.created",
+            "response": {"id": "resp-1"}
+        }),
+        ev_function_call(call_id, "list_dir", &arguments),
+        ev_completed("resp-1"),
+    ]);
+    responses::mount_sse_once_match(&server, any(), first_response).await;
+
+    let second_response = sse(vec![
+        ev_assistant_message("msg-1", "done"),
+        ev_completed("resp-2"),
+    ]);
+    responses::mount_sse_once_match(&server, any(), second_response).await;
+
+    let session_model = session_configured.model.clone();
+
+    codex
+        .submit(Op::UserTurn {
+            items: vec![InputItem::Text {
+                text: "list directory contents depth two".into(),
+            }],
+            final_output_json_schema: None,
+            cwd: cwd.path().to_path_buf(),
+            approval_policy: AskForApproval::Never,
+            sandbox_policy: SandboxPolicy::DangerFullAccess,
+            model: session_model,
+            effort: None,
+            summary: ReasoningSummary::Auto,
+        })
+        .await?;
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let requests = server.received_requests().await.expect("recorded requests");
+    let request_bodies = requests
+        .iter()
+        .map(|req| req.body_json::<Value>().unwrap())
+        .collect::<Vec<_>>();
+    assert!(
+        !request_bodies.is_empty(),
+        "expected at least one request body"
+    );
+
+    let tool_output_item = request_bodies
+        .iter()
+        .find_map(|body| {
+            body.get("input")
+                .and_then(Value::as_array)
+                .and_then(|items| {
+                    items.iter().find(|item| {
+                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
+                    })
+                })
+        })
+        .unwrap_or_else(|| {
+            panic!("function_call_output item not found in requests: {request_bodies:#?}")
+        });
+
+    assert_eq!(
+        tool_output_item.get("call_id").and_then(Value::as_str),
+        Some(call_id)
+    );
+
+    let output_text = tool_output_item
+        .get("output")
+        .and_then(|value| match value {
+            Value::String(text) => Some(text.as_str()),
+            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
+            _ => None,
+        })
+        .expect("output text present");
+    assert_eq!(
+        output_text,
+        "E1: [file] alpha.txt\nE2: [dir] nested\nE3: [file] nested/beta.txt\nE4: [dir] nested/grand"
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+#[ignore = "disabled until we enable list_dir tool"]
+async fn list_dir_tool_depth_three_includes_grandchildren() -> anyhow::Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+
+    let TestCodex {
+        codex,
+        cwd,
+        session_configured,
+        ..
+    } = test_codex().build(&server).await?;
+
+    let dir_path = cwd.path().join("depth_three");
+    std::fs::create_dir(&dir_path)?;
+    std::fs::write(dir_path.join("alpha.txt"), "alpha")?;
+    let nested = dir_path.join("nested");
+    std::fs::create_dir(&nested)?;
+    std::fs::write(nested.join("beta.txt"), "beta")?;
+    let deeper = nested.join("grand");
+    std::fs::create_dir(&deeper)?;
+    std::fs::write(deeper.join("gamma.txt"), "gamma")?;
+    let dir_path_string = dir_path.to_string_lossy().to_string();
+
+    let call_id = "list-dir-depth3";
+    let arguments = serde_json::json!({
+        "dir_path": dir_path_string,
+        "offset": 1,
+        "limit": 10,
+        "depth": 3,
+    })
+    .to_string();
+
+    let first_response = sse(vec![
+        serde_json::json!({
+            "type": "response.created",
+            "response": {"id": "resp-1"}
+        }),
+        ev_function_call(call_id, "list_dir", &arguments),
+        ev_completed("resp-1"),
+    ]);
+    responses::mount_sse_once_match(&server, any(), first_response).await;
+
+    let second_response = sse(vec![
+        ev_assistant_message("msg-1", "done"),
+        ev_completed("resp-2"),
+    ]);
+    responses::mount_sse_once_match(&server, any(), second_response).await;
+
+    let session_model = session_configured.model.clone();
+
+    codex
+        .submit(Op::UserTurn {
+            items: vec![InputItem::Text {
+                text: "list directory contents depth three".into(),
+            }],
+            final_output_json_schema: None,
+            cwd: cwd.path().to_path_buf(),
+            approval_policy: AskForApproval::Never,
+            sandbox_policy: SandboxPolicy::DangerFullAccess,
+            model: session_model,
+            effort: None,
+            summary: ReasoningSummary::Auto,
+        })
+        .await?;
+
+    wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
+
+    let requests = server.received_requests().await.expect("recorded requests");
+    let request_bodies = requests
+        .iter()
+        .map(|req| req.body_json::<Value>().unwrap())
+        .collect::<Vec<_>>();
+    assert!(
+        !request_bodies.is_empty(),
+        "expected at least one request body"
+    );
+
+    let tool_output_item = request_bodies
+        .iter()
+        .find_map(|body| {
+            body.get("input")
+                .and_then(Value::as_array)
+                .and_then(|items| {
+                    items.iter().find(|item| {
+                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
+                    })
+                })
+        })
+        .unwrap_or_else(|| {
+            panic!("function_call_output item not found in requests: {request_bodies:#?}")
+        });
+
+    assert_eq!(
+        tool_output_item.get("call_id").and_then(Value::as_str),
+        Some(call_id)
+    );
+
+    let output_text = tool_output_item
+        .get("output")
+        .and_then(|value| match value {
+            Value::String(text) => Some(text.as_str()),
+            Value::Object(obj) => obj.get("content").and_then(Value::as_str),
+            _ => None,
+        })
+        .expect("output text present");
+    assert_eq!(
+        output_text,
+        "E1: [file] alpha.txt\nE2: [dir] nested\nE3: [file] nested/beta.txt\nE4: [dir] nested/grand\nE5: [file] nested/grand/gamma.txt"
+    );
+
+    Ok(())
+}

codex-rs/core/tests/suite/mod.rs

@@ -9,7 +9,9 @@ mod compact_resume_fork;
 mod exec;
 mod exec_stream_events;
 mod fork_conversation;
+mod grep_files;
 mod json_result;
+mod list_dir;
 mod live_cli;
 mod model_overrides;
 mod model_tools;

codex-rs/core/tests/suite/model_tools.rs

@@ -4,20 +4,18 @@ use codex_core::CodexAuth;
 use codex_core::ConversationManager;
 use codex_core::ModelProviderInfo;
 use codex_core::built_in_model_providers;
+use codex_core::features::Feature;
 use codex_core::model_family::find_family_for_model;
 use codex_core::protocol::EventMsg;
 use codex_core::protocol::InputItem;
 use codex_core::protocol::Op;
 use core_test_support::load_default_config_for_test;
 use core_test_support::load_sse_fixture_with_id;
+use core_test_support::responses;
 use core_test_support::skip_if_no_network;
 use core_test_support::wait_for_event;
 use tempfile::TempDir;
-use wiremock::Mock;
 use wiremock::MockServer;
-use wiremock::ResponseTemplate;
-use wiremock::matchers::method;
-use wiremock::matchers::path;
 
 fn sse_completed(id: &str) -> String {
     load_sse_fixture_with_id("tests/fixtures/completed_template.json", id)
@@ -44,16 +42,7 @@ async fn collect_tool_identifiers_for_model(model: &str) -> Vec<String> {
     let server = MockServer::start().await;
 
     let sse = sse_completed(model);
-    let template = ResponseTemplate::new(200)
-        .insert_header("content-type", "text/event-stream")
-        .set_body_raw(sse, "text/event-stream");
-
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .respond_with(template)
-        .expect(1)
-        .mount(&server)
-        .await;
+    let resp_mock = responses::mount_sse_once_match(&server, wiremock::matchers::any(), sse).await;
 
     let model_provider = ModelProviderInfo {
         base_url: Some(format!("{}/v1", server.uri())),
@@ -68,12 +57,12 @@ async fn collect_tool_identifiers_for_model(model: &str) -> Vec<String> {
     config.model = model.to_string();
     config.model_family =
         find_family_for_model(model).unwrap_or_else(|| panic!("unknown model family for {model}"));
-    config.include_plan_tool = false;
-    config.include_apply_patch_tool = false;
-    config.include_view_image_tool = false;
-    config.tools_web_search_request = false;
-    config.use_experimental_streamable_shell_tool = false;
-    config.use_experimental_unified_exec_tool = false;
+    config.features.disable(Feature::PlanTool);
+    config.features.disable(Feature::ApplyPatchFreeform);
+    config.features.disable(Feature::ViewImageTool);
+    config.features.disable(Feature::WebSearchRequest);
+    config.features.disable(Feature::StreamableShell);
+    config.features.disable(Feature::UnifiedExec);
 
     let conversation_manager =
         ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
@@ -93,13 +82,7 @@ async fn collect_tool_identifiers_for_model(model: &str) -> Vec<String> {
         .unwrap();
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.unwrap();
-    assert_eq!(
-        requests.len(),
-        1,
-        "expected a single request for model {model}"
-    );
-    let body = requests[0].body_json::<serde_json::Value>().unwrap();
+    let body = resp_mock.single_request().body_json();
     tool_identifiers(&body)
 }
 

codex-rs/core/tests/suite/prompt_caching.rs

@@ -5,6 +5,7 @@ use codex_core::ConversationManager;
 use codex_core::ModelProviderInfo;
 use codex_core::built_in_model_providers;
 use codex_core::config::OPENAI_DEFAULT_MODEL;
+use codex_core::features::Feature;
 use codex_core::model_family::find_family_for_model;
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::EventMsg;
@@ -99,10 +100,10 @@ async fn codex_mini_latest_tools() {
     config.cwd = cwd.path().to_path_buf();
     config.model_provider = model_provider;
     config.user_instructions = Some("be consistent and helpful".to_string());
+    config.features.disable(Feature::ApplyPatchFreeform);
 
     let conversation_manager =
         ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));
-    config.include_apply_patch_tool = false;
     config.model = "codex-mini-latest".to_string();
     config.model_family = find_family_for_model("codex-mini-latest").unwrap();
 
@@ -185,7 +186,7 @@ async fn prompt_tools_are_consistent_across_requests() {
     config.cwd = cwd.path().to_path_buf();
     config.model_provider = model_provider;
     config.user_instructions = Some("be consistent and helpful".to_string());
-    config.include_plan_tool = true;
+    config.features.enable(Feature::PlanTool);
 
     let conversation_manager =
         ConversationManager::with_auth(CodexAuth::from_api_key("Test API Key"));

codex-rs/core/tests/suite/read_file.rs

@@ -10,6 +10,7 @@ use core_test_support::responses;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_function_call;
+use core_test_support::responses::ev_response_created;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
@@ -47,10 +48,7 @@ async fn read_file_tool_returns_requested_lines() -> anyhow::Result<()> {
     .to_string();
 
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
+        ev_response_created("resp-1"),
         ev_function_call(call_id, "read_file", &arguments),
         ev_completed("resp-1"),
     ]);
@@ -60,7 +58,7 @@ async fn read_file_tool_returns_requested_lines() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -81,36 +79,12 @@ async fn read_file_tool_returns_requested_lines() -> anyhow::Result<()> {
 
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().unwrap())
-        .collect::<Vec<_>>();
-    assert!(
-        !request_bodies.is_empty(),
-        "expected at least one request body"
-    );
-
-    let tool_output_item = request_bodies
-        .iter()
-        .find_map(|body| {
-            body.get("input")
-                .and_then(Value::as_array)
-                .and_then(|items| {
-                    items.iter().find(|item| {
-                        item.get("type").and_then(Value::as_str) == Some("function_call_output")
-                    })
-                })
-        })
-        .unwrap_or_else(|| {
-            panic!("function_call_output item not found in requests: {request_bodies:#?}")
-        });
-
+    let req = second_mock.single_request();
+    let tool_output_item = req.function_call_output(call_id);
     assert_eq!(
         tool_output_item.get("call_id").and_then(Value::as_str),
         Some(call_id)
     );
-
     let output_text = tool_output_item
         .get("output")
         .and_then(|value| match value {

codex-rs/core/tests/suite/review.rs

@@ -445,7 +445,7 @@ async fn review_input_isolated_from_parent_history() {
     .await;
     let _complete = wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
-    // Assert the request `input` contains the environment context followed by the review prompt.
+    // Assert the request `input` contains the environment context followed by the user review prompt.
     let request = &server.received_requests().await.unwrap()[0];
     let body = request.body_json::<serde_json::Value>().unwrap();
     let input = body["input"].as_array().expect("input array");
@@ -473,9 +473,14 @@ async fn review_input_isolated_from_parent_history() {
     assert_eq!(review_msg["role"].as_str().unwrap(), "user");
     assert_eq!(
         review_msg["content"][0]["text"].as_str().unwrap(),
-        format!("{REVIEW_PROMPT}\n\n---\n\nNow, here's your task: Please review only this",)
+        review_prompt,
+        "user message should only contain the raw review prompt"
     );
 
+    // Ensure the REVIEW_PROMPT rubric is sent via instructions.
+    let instructions = body["instructions"].as_str().expect("instructions string");
+    assert_eq!(instructions, REVIEW_PROMPT);
+
     // Also verify that a user interruption note was recorded in the rollout.
     codex.submit(Op::GetPath).await.unwrap();
     let history_event =

codex-rs/core/tests/suite/rmcp_client.rs

@@ -9,6 +9,7 @@ use std::time::UNIX_EPOCH;
 
 use codex_core::config_types::McpServerConfig;
 use codex_core::config_types::McpServerTransportConfig;
+use codex_core::features::Feature;
 
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::EventMsg;
@@ -47,10 +48,7 @@ async fn stdio_server_round_trip() -> anyhow::Result<()> {
         &server,
         any(),
         responses::sse(vec![
-            serde_json::json!({
-                "type": "response.created",
-                "response": {"id": "resp-1"}
-            }),
+            responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, "{\"message\":\"ping\"}"),
             responses::ev_completed("resp-1"),
         ]),
@@ -77,7 +75,7 @@ async fn stdio_server_round_trip() -> anyhow::Result<()> {
 
     let fixture = test_codex()
         .with_config(move |config| {
-            config.use_experimental_use_rmcp_client = true;
+            config.features.enable(Feature::RmcpClient);
             config.mcp_servers.insert(
                 server_name.to_string(),
                 McpServerConfig {
@@ -89,6 +87,7 @@ async fn stdio_server_round_trip() -> anyhow::Result<()> {
                             expected_env_value.to_string(),
                         )])),
                     },
+                    enabled: true,
                     startup_timeout_sec: Some(Duration::from_secs(10)),
                     tool_timeout_sec: None,
                 },
@@ -184,10 +183,7 @@ async fn streamable_http_tool_call_round_trip() -> anyhow::Result<()> {
         &server,
         any(),
         responses::sse(vec![
-            serde_json::json!({
-                "type": "response.created",
-                "response": {"id": "resp-1"}
-            }),
+            responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, "{\"message\":\"ping\"}"),
             responses::ev_completed("resp-1"),
         ]),
@@ -232,14 +228,15 @@ async fn streamable_http_tool_call_round_trip() -> anyhow::Result<()> {
 
     let fixture = test_codex()
         .with_config(move |config| {
-            config.use_experimental_use_rmcp_client = true;
+            config.features.enable(Feature::RmcpClient);
             config.mcp_servers.insert(
                 server_name.to_string(),
                 McpServerConfig {
                     transport: McpServerTransportConfig::StreamableHttp {
                         url: server_url,
-                        bearer_token: None,
+                        bearer_token_env_var: None,
                     },
+                    enabled: true,
                     startup_timeout_sec: Some(Duration::from_secs(10)),
                     tool_timeout_sec: None,
                 },
@@ -352,10 +349,7 @@ async fn streamable_http_with_oauth_round_trip() -> anyhow::Result<()> {
         &server,
         any(),
         responses::sse(vec![
-            serde_json::json!({
-                "type": "response.created",
-                "response": {"id": "resp-1"}
-            }),
+            responses::ev_response_created("resp-1"),
             responses::ev_function_call(call_id, &tool_name, "{\"message\":\"ping\"}"),
             responses::ev_completed("resp-1"),
         ]),
@@ -415,14 +409,15 @@ async fn streamable_http_with_oauth_round_trip() -> anyhow::Result<()> {
 
     let fixture = test_codex()
         .with_config(move |config| {
-            config.use_experimental_use_rmcp_client = true;
+            config.features.enable(Feature::RmcpClient);
             config.mcp_servers.insert(
                 server_name.to_string(),
                 McpServerConfig {
                     transport: McpServerTransportConfig::StreamableHttp {
                         url: server_url,
-                        bearer_token: None,
+                        bearer_token_env_var: None,
                     },
+                    enabled: true,
                     startup_timeout_sec: Some(Duration::from_secs(10)),
                     tool_timeout_sec: None,
                 },

codex-rs/core/tests/suite/shell_serialization.rs

@@ -1,6 +1,7 @@
 #![cfg(not(target_os = "windows"))]
 
 use anyhow::Result;
+use codex_core::features::Feature;
 use codex_core::model_family::find_family_for_model;
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::EventMsg;
@@ -8,9 +9,14 @@ use codex_core::protocol::InputItem;
 use codex_core::protocol::Op;
 use codex_core::protocol::SandboxPolicy;
 use codex_protocol::config_types::ReasoningSummary;
+use core_test_support::assert_regex_match;
+use core_test_support::responses::ev_apply_patch_function_call;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
+use core_test_support::responses::ev_custom_tool_call;
 use core_test_support::responses::ev_function_call;
+use core_test_support::responses::ev_local_shell_call;
+use core_test_support::responses::ev_response_created;
 use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
@@ -18,8 +24,11 @@ use core_test_support::skip_if_no_network;
 use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event;
+use pretty_assertions::assert_eq;
+use regex_lite::Regex;
 use serde_json::Value;
 use serde_json::json;
+use std::fs;
 
 async fn submit_turn(test: &TestCodex, prompt: &str, sandbox_policy: SandboxPolicy) -> Result<()> {
     let session_model = test.session_configured.model.clone();
@@ -69,13 +78,28 @@ fn find_function_call_output<'a>(bodies: &'a [Value], call_id: &str) -> Option<&
     None
 }
 
+fn find_custom_tool_call_output<'a>(bodies: &'a [Value], call_id: &str) -> Option<&'a Value> {
+    for body in bodies {
+        if let Some(items) = body.get("input").and_then(Value::as_array) {
+            for item in items {
+                if item.get("type").and_then(Value::as_str) == Some("custom_tool_call_output")
+                    && item.get("call_id").and_then(Value::as_str) == Some(call_id)
+                {
+                    return Some(item);
+                }
+            }
+        }
+    }
+    None
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn shell_output_stays_json_without_freeform_apply_patch() -> Result<()> {
     skip_if_no_network!(Ok(()));
 
     let server = start_mock_server().await;
     let mut builder = test_codex().with_config(|config| {
-        config.include_apply_patch_tool = false;
+        config.features.disable(Feature::ApplyPatchFreeform);
         config.model = "gpt-5".to_string();
         config.model_family = find_family_for_model("gpt-5").expect("gpt-5 is a model family");
     });
@@ -88,7 +112,7 @@ async fn shell_output_stays_json_without_freeform_apply_patch() -> Result<()> {
     });
     let responses = vec![
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
             ev_completed("resp-1"),
         ]),
@@ -117,7 +141,12 @@ async fn shell_output_stays_json_without_freeform_apply_patch() -> Result<()> {
         .and_then(Value::as_str)
         .expect("shell output string");
 
-    let parsed: Value = serde_json::from_str(output)?;
+    let mut parsed: Value = serde_json::from_str(output)?;
+    if let Some(metadata) = parsed.get_mut("metadata").and_then(Value::as_object_mut) {
+        // duration_seconds is non-deterministic; remove it for deep equality
+        let _ = metadata.remove("duration_seconds");
+    }
+
     assert_eq!(
         parsed
             .get("metadata")
@@ -130,10 +159,7 @@ async fn shell_output_stays_json_without_freeform_apply_patch() -> Result<()> {
         .get("output")
         .and_then(Value::as_str)
         .unwrap_or_default();
-    assert!(
-        stdout.contains("shell json"),
-        "expected stdout to include command output, got {stdout:?}"
-    );
+    assert_regex_match(r"(?s)^shell json\n?$", stdout);
 
     Ok(())
 }
@@ -144,7 +170,7 @@ async fn shell_output_is_structured_with_freeform_apply_patch() -> Result<()> {
 
     let server = start_mock_server().await;
     let mut builder = test_codex().with_config(|config| {
-        config.include_apply_patch_tool = true;
+        config.features.enable(Feature::ApplyPatchFreeform);
     });
     let test = builder.build(&server).await?;
 
@@ -155,7 +181,7 @@ async fn shell_output_is_structured_with_freeform_apply_patch() -> Result<()> {
     });
     let responses = vec![
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
             ev_completed("resp-1"),
         ]),
@@ -189,17 +215,88 @@ async fn shell_output_is_structured_with_freeform_apply_patch() -> Result<()> {
         serde_json::from_str::<Value>(output).is_err(),
         "expected structured shell output to be plain text",
     );
+    let expected_pattern = r"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+freeform shell
+?$";
+    assert_regex_match(expected_pattern, output);
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn shell_output_for_freeform_tool_records_duration() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.include_apply_patch_tool = true;
+    });
+    let test = builder.build(&server).await?;
+
+    #[cfg(target_os = "linux")]
+    let sleep_cmd = vec!["/bin/bash", "-c", "sleep 1"];
+
+    #[cfg(target_os = "macos")]
+    let sleep_cmd = vec!["/bin/bash", "-c", "sleep 1"];
+
+    #[cfg(windows)]
+    let sleep_cmd = "timeout 1";
+
+    let call_id = "shell-structured";
+    let args = json!({
+        "command": sleep_cmd,
+        "timeout_ms": 2_000,
+    });
+    let responses = vec![
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "done"),
+            ev_completed("resp-2"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    submit_turn(
+        &test,
+        "run the structured shell command",
+        SandboxPolicy::DangerFullAccess,
+    )
+    .await?;
+
+    let requests = server
+        .received_requests()
+        .await
+        .expect("recorded requests present");
+    let bodies = request_bodies(&requests)?;
+    let output_item =
+        find_function_call_output(&bodies, call_id).expect("structured output present");
+    let output = output_item
+        .get("output")
+        .and_then(Value::as_str)
+        .expect("structured output string");
+
+    let expected_pattern = r#"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+$"#;
+    assert_regex_match(expected_pattern, output);
+
+    let wall_time_regex = Regex::new(r"(?m)^Wall (?:time|Clock): ([0-9]+(?:\.[0-9]+)?) seconds$")
+        .expect("compile wall time regex");
+    let wall_time_seconds = wall_time_regex
+        .captures(output)
+        .and_then(|caps| caps.get(1))
+        .and_then(|value| value.as_str().parse::<f32>().ok())
+        .expect("expected structured shell output to contain wall time seconds");
     assert!(
-        output.starts_with("Exit code: 0\n"),
-        "expected exit code prefix, got {output:?}",
-    );
-    assert!(
-        output.contains("\nOutput:\n"),
-        "expected Output section, got {output:?}"
-    );
-    assert!(
-        output.contains("freeform shell"),
-        "expected stdout content, got {output:?}"
+        wall_time_seconds > 0.5,
+        "expected wall time to be greater than zero seconds, got {wall_time_seconds}"
     );
 
     Ok(())
@@ -220,11 +317,11 @@ async fn shell_output_reserializes_truncated_content() -> Result<()> {
     let call_id = "shell-truncated";
     let args = json!({
         "command": ["/bin/sh", "-c", "seq 1 400"],
-        "timeout_ms": 1_000,
+        "timeout_ms": 5_000,
     });
     let responses = vec![
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
             ev_completed("resp-1"),
         ]),
@@ -258,18 +355,452 @@ async fn shell_output_reserializes_truncated_content() -> Result<()> {
         serde_json::from_str::<Value>(output).is_err(),
         "expected truncated shell output to be plain text",
     );
-    assert!(
-        output.starts_with("Exit code: 0\n"),
-        "expected exit code prefix, got {output:?}",
+    let truncated_pattern = r#"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Total output lines: 400
+Output:
+1
+2
+3
+4
+5
+6
+.*
+\[\.{3} omitted \d+ of 400 lines \.{3}\]
+
+.*
+396
+397
+398
+399
+400
+$"#;
+    assert_regex_match(truncated_pattern, output);
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn apply_patch_custom_tool_output_is_structured() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.include_apply_patch_tool = true;
+    });
+    let test = builder.build(&server).await?;
+
+    let call_id = "apply-patch-structured";
+    let file_name = "structured.txt";
+    let patch = format!(
+        r#"*** Begin Patch
+*** Add File: {file_name}
++from custom tool
+*** End Patch
+"#
     );
-    assert!(
-        output.lines().any(|line| line == "Total output lines: 400"),
-        "expected total output lines marker, got {output:?}",
+    let responses = vec![
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_custom_tool_call(call_id, "apply_patch", &patch),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "done"),
+            ev_completed("resp-2"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    submit_turn(
+        &test,
+        "apply the patch via custom tool",
+        SandboxPolicy::DangerFullAccess,
+    )
+    .await?;
+
+    let requests = server
+        .received_requests()
+        .await
+        .expect("recorded requests present");
+    let bodies = request_bodies(&requests)?;
+    let output_item =
+        find_custom_tool_call_output(&bodies, call_id).expect("apply_patch output present");
+    let output = output_item
+        .get("output")
+        .and_then(Value::as_str)
+        .expect("apply_patch output string");
+
+    let expected_pattern = format!(
+        r"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+Success. Updated the following files:
+A {file_name}
+?$"
     );
-    assert!(
-        output.contains("[... omitted"),
-        "expected truncated marker, got {output:?}",
+    assert_regex_match(&expected_pattern, output);
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn apply_patch_custom_tool_call_creates_file() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.include_apply_patch_tool = true;
+    });
+    let test = builder.build(&server).await?;
+
+    let call_id = "apply-patch-add-file";
+    let file_name = "custom_tool_apply_patch.txt";
+    let patch = format!(
+        "*** Begin Patch\n*** Add File: {file_name}\n+custom tool content\n*** End Patch\n"
+    );
+    let responses = vec![
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_custom_tool_call(call_id, "apply_patch", &patch),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "apply_patch done"),
+            ev_completed("resp-2"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    submit_turn(
+        &test,
+        "apply the patch via custom tool to create a file",
+        SandboxPolicy::DangerFullAccess,
+    )
+    .await?;
+
+    let requests = server
+        .received_requests()
+        .await
+        .expect("recorded requests present");
+    let bodies = request_bodies(&requests)?;
+    let output_item =
+        find_custom_tool_call_output(&bodies, call_id).expect("apply_patch output present");
+    let output = output_item
+        .get("output")
+        .and_then(Value::as_str)
+        .expect("apply_patch output string");
+
+    let expected_pattern = format!(
+        r"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+Success. Updated the following files:
+A {file_name}
+?$"
+    );
+    assert_regex_match(&expected_pattern, output);
+
+    let new_file_path = test.cwd.path().join(file_name);
+    let created_contents = fs::read_to_string(&new_file_path)?;
+    assert_eq!(
+        created_contents, "custom tool content\n",
+        "expected file contents for {file_name}"
     );
 
     Ok(())
 }
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn apply_patch_custom_tool_call_updates_existing_file() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.include_apply_patch_tool = true;
+    });
+    let test = builder.build(&server).await?;
+
+    let call_id = "apply-patch-update-file";
+    let file_name = "custom_tool_apply_patch_existing.txt";
+    let file_path = test.cwd.path().join(file_name);
+    fs::write(&file_path, "before\n")?;
+    let patch = format!(
+        "*** Begin Patch\n*** Update File: {file_name}\n@@\n-before\n+after\n*** End Patch\n"
+    );
+    let responses = vec![
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_custom_tool_call(call_id, "apply_patch", &patch),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "apply_patch update done"),
+            ev_completed("resp-2"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    submit_turn(
+        &test,
+        "apply the patch via custom tool to update a file",
+        SandboxPolicy::DangerFullAccess,
+    )
+    .await?;
+
+    let requests = server
+        .received_requests()
+        .await
+        .expect("recorded requests present");
+    let bodies = request_bodies(&requests)?;
+    let output_item =
+        find_custom_tool_call_output(&bodies, call_id).expect("apply_patch output present");
+    let output = output_item
+        .get("output")
+        .and_then(Value::as_str)
+        .expect("apply_patch output string");
+
+    let expected_pattern = format!(
+        r"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+Success. Updated the following files:
+M {file_name}
+?$"
+    );
+    assert_regex_match(&expected_pattern, output);
+
+    let updated_contents = fs::read_to_string(file_path)?;
+    assert_eq!(updated_contents, "after\n", "expected updated file content");
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn apply_patch_custom_tool_call_reports_failure_output() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.include_apply_patch_tool = true;
+    });
+    let test = builder.build(&server).await?;
+
+    let call_id = "apply-patch-failure";
+    let missing_file = "missing_custom_tool_apply_patch.txt";
+    let patch = format!(
+        "*** Begin Patch\n*** Update File: {missing_file}\n@@\n-before\n+after\n*** End Patch\n"
+    );
+    let responses = vec![
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_custom_tool_call(call_id, "apply_patch", &patch),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "apply_patch failure done"),
+            ev_completed("resp-2"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    submit_turn(
+        &test,
+        "attempt a failing apply_patch via custom tool",
+        SandboxPolicy::DangerFullAccess,
+    )
+    .await?;
+
+    let requests = server
+        .received_requests()
+        .await
+        .expect("recorded requests present");
+    let bodies = request_bodies(&requests)?;
+    let output_item =
+        find_custom_tool_call_output(&bodies, call_id).expect("apply_patch output present");
+    let output = output_item
+        .get("output")
+        .and_then(Value::as_str)
+        .expect("apply_patch output string");
+
+    let expected_output = format!(
+        "apply_patch verification failed: Failed to read file to update {}/{missing_file}: No such file or directory (os error 2)",
+        test.cwd.path().to_string_lossy()
+    );
+    assert_eq!(output, expected_output);
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn apply_patch_function_call_output_is_structured() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.include_apply_patch_tool = true;
+    });
+    let test = builder.build(&server).await?;
+
+    let call_id = "apply-patch-function";
+    let file_name = "function_apply_patch.txt";
+    let patch =
+        format!("*** Begin Patch\n*** Add File: {file_name}\n+via function call\n*** End Patch\n");
+    let responses = vec![
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_apply_patch_function_call(call_id, &patch),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "apply_patch function done"),
+            ev_completed("resp-2"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    submit_turn(
+        &test,
+        "apply the patch via function-call apply_patch",
+        SandboxPolicy::DangerFullAccess,
+    )
+    .await?;
+
+    let requests = server
+        .received_requests()
+        .await
+        .expect("recorded requests present");
+    let bodies = request_bodies(&requests)?;
+    let output_item =
+        find_function_call_output(&bodies, call_id).expect("apply_patch function output present");
+    let output = output_item
+        .get("output")
+        .and_then(Value::as_str)
+        .expect("apply_patch output string");
+
+    let expected_pattern = format!(
+        r"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+Success. Updated the following files:
+A {file_name}
+?$"
+    );
+    assert_regex_match(&expected_pattern, output);
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn shell_output_is_structured_for_nonzero_exit() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.model = "gpt-5-codex".to_string();
+        config.model_family =
+            find_family_for_model("gpt-5-codex").expect("gpt-5-codex is a model family");
+        config.include_apply_patch_tool = true;
+    });
+    let test = builder.build(&server).await?;
+
+    let call_id = "shell-nonzero-exit";
+    let args = json!({
+        "command": ["/bin/sh", "-c", "exit 42"],
+        "timeout_ms": 1_000,
+    });
+    let responses = vec![
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "shell failure handled"),
+            ev_completed("resp-2"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    submit_turn(
+        &test,
+        "run the failing shell command",
+        SandboxPolicy::DangerFullAccess,
+    )
+    .await?;
+
+    let requests = server
+        .received_requests()
+        .await
+        .expect("recorded requests present");
+    let bodies = request_bodies(&requests)?;
+    let output_item = find_function_call_output(&bodies, call_id).expect("shell output present");
+    let output = output_item
+        .get("output")
+        .and_then(Value::as_str)
+        .expect("shell output string");
+
+    let expected_pattern = r"(?s)^Exit code: 42
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+?$";
+    assert_regex_match(expected_pattern, output);
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn local_shell_call_output_is_structured() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+
+    let server = start_mock_server().await;
+    let mut builder = test_codex().with_config(|config| {
+        config.model = "gpt-5-codex".to_string();
+        config.model_family =
+            find_family_for_model("gpt-5-codex").expect("gpt-5-codex is a model family");
+        config.include_apply_patch_tool = true;
+    });
+    let test = builder.build(&server).await?;
+
+    let call_id = "local-shell-call";
+    let responses = vec![
+        sse(vec![
+            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_local_shell_call(call_id, "completed", vec!["/bin/echo", "local shell"]),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "local shell done"),
+            ev_completed("resp-2"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    submit_turn(
+        &test,
+        "run the local shell command",
+        SandboxPolicy::DangerFullAccess,
+    )
+    .await?;
+
+    let requests = server
+        .received_requests()
+        .await
+        .expect("recorded requests present");
+    let bodies = request_bodies(&requests)?;
+    let output_item =
+        find_function_call_output(&bodies, call_id).expect("local shell output present");
+    let output = output_item
+        .get("output")
+        .and_then(Value::as_str)
+        .expect("local shell output string");
+
+    let expected_pattern = r"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+local shell
+?$";
+    assert_regex_match(expected_pattern, output);
+
+    Ok(())
+}

codex-rs/core/tests/suite/tool_harness.rs

@@ -1,5 +1,9 @@
 #![cfg(not(target_os = "windows"))]
 
+use std::fs;
+
+use assert_matches::assert_matches;
+use codex_core::features::Feature;
 use codex_core::model_family::find_family_for_model;
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::EventMsg;
@@ -8,12 +12,14 @@ use codex_core::protocol::Op;
 use codex_core::protocol::SandboxPolicy;
 use codex_protocol::config_types::ReasoningSummary;
 use codex_protocol::plan_tool::StepStatus;
+use core_test_support::assert_regex_match;
 use core_test_support::responses;
 use core_test_support::responses::ev_apply_patch_function_call;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_function_call;
 use core_test_support::responses::ev_local_shell_call;
+use core_test_support::responses::ev_response_created;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
@@ -24,16 +30,6 @@ use serde_json::Value;
 use serde_json::json;
 use wiremock::matchers::any;
 
-fn function_call_output(body: &Value) -> Option<&Value> {
-    body.get("input")
-        .and_then(Value::as_array)
-        .and_then(|items| {
-            items.iter().find(|item| {
-                item.get("type").and_then(Value::as_str) == Some("function_call_output")
-            })
-        })
-}
-
 fn extract_output_text(item: &Value) -> Option<&str> {
     item.get("output").and_then(|value| match value {
         Value::String(text) => Some(text.as_str()),
@@ -42,12 +38,6 @@ fn extract_output_text(item: &Value) -> Option<&str> {
     })
 }
 
-fn find_request_with_function_call_output(requests: &[Value]) -> Option<&Value> {
-    requests
-        .iter()
-        .find(|body| function_call_output(body).is_some())
-}
-
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn shell_tool_executes_command_and_streams_output() -> anyhow::Result<()> {
     skip_if_no_network!(Ok(()));
@@ -68,10 +58,7 @@ async fn shell_tool_executes_command_and_streams_output() -> anyhow::Result<()>
     let call_id = "shell-tool-call";
     let command = vec!["/bin/echo", "tool harness"];
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
+        ev_response_created("resp-1"),
         ev_local_shell_call(call_id, "completed", command),
         ev_completed("resp-1"),
     ]);
@@ -81,7 +68,7 @@ async fn shell_tool_executes_command_and_streams_output() -> anyhow::Result<()>
         ev_assistant_message("msg-1", "all done"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -102,25 +89,13 @@ async fn shell_tool_executes_command_and_streams_output() -> anyhow::Result<()>
 
     wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert!(!requests.is_empty(), "expected at least one POST request");
-
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect::<Vec<_>>();
-
-    let body_with_tool_output = find_request_with_function_call_output(&request_bodies)
-        .expect("function_call_output item not found in requests");
-    let output_item = function_call_output(body_with_tool_output).expect("tool output item");
-    let output_text = extract_output_text(output_item).expect("output text present");
+    let req = second_mock.single_request();
+    let output_item = req.function_call_output(call_id);
+    let output_text = extract_output_text(&output_item).expect("output text present");
     let exec_output: Value = serde_json::from_str(output_text)?;
     assert_eq!(exec_output["metadata"]["exit_code"], 0);
     let stdout = exec_output["output"].as_str().expect("stdout field");
-    assert!(
-        stdout.contains("tool harness"),
-        "expected stdout to contain command output, got {stdout:?}"
-    );
+    assert_regex_match(r"(?s)^tool harness\n?$", stdout);
 
     Ok(())
 }
@@ -132,7 +107,7 @@ async fn update_plan_tool_emits_plan_update_event() -> anyhow::Result<()> {
     let server = start_mock_server().await;
 
     let mut builder = test_codex().with_config(|config| {
-        config.include_plan_tool = true;
+        config.features.enable(Feature::PlanTool);
     });
     let TestCodex {
         codex,
@@ -152,10 +127,7 @@ async fn update_plan_tool_emits_plan_update_event() -> anyhow::Result<()> {
     .to_string();
 
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
+        ev_response_created("resp-1"),
         ev_function_call(call_id, "update_plan", &plan_args),
         ev_completed("resp-1"),
     ]);
@@ -165,7 +137,7 @@ async fn update_plan_tool_emits_plan_update_event() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "plan acknowledged"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -191,9 +163,9 @@ async fn update_plan_tool_emits_plan_update_event() -> anyhow::Result<()> {
             assert_eq!(update.explanation.as_deref(), Some("Tool harness check"));
             assert_eq!(update.plan.len(), 2);
             assert_eq!(update.plan[0].step, "Inspect workspace");
-            assert!(matches!(update.plan[0].status, StepStatus::InProgress));
+            assert_matches!(update.plan[0].status, StepStatus::InProgress);
             assert_eq!(update.plan[1].step, "Report results");
-            assert!(matches!(update.plan[1].status, StepStatus::Pending));
+            assert_matches!(update.plan[1].status, StepStatus::Pending);
             false
         }
         EventMsg::TaskComplete(_) => true,
@@ -203,22 +175,13 @@ async fn update_plan_tool_emits_plan_update_event() -> anyhow::Result<()> {
 
     assert!(saw_plan_update, "expected PlanUpdate event");
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert!(!requests.is_empty(), "expected at least one POST request");
-
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect::<Vec<_>>();
-
-    let body_with_tool_output = find_request_with_function_call_output(&request_bodies)
-        .expect("function_call_output item not found in requests");
-    let output_item = function_call_output(body_with_tool_output).expect("tool output item");
+    let req = second_mock.single_request();
+    let output_item = req.function_call_output(call_id);
     assert_eq!(
         output_item.get("call_id").and_then(Value::as_str),
         Some(call_id)
     );
-    let output_text = extract_output_text(output_item).expect("output text present");
+    let output_text = extract_output_text(&output_item).expect("output text present");
     assert_eq!(output_text, "Plan updated");
 
     Ok(())
@@ -231,7 +194,7 @@ async fn update_plan_tool_rejects_malformed_payload() -> anyhow::Result<()> {
     let server = start_mock_server().await;
 
     let mut builder = test_codex().with_config(|config| {
-        config.include_plan_tool = true;
+        config.features.enable(Feature::PlanTool);
     });
     let TestCodex {
         codex,
@@ -247,10 +210,7 @@ async fn update_plan_tool_rejects_malformed_payload() -> anyhow::Result<()> {
     .to_string();
 
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
+        ev_response_created("resp-1"),
         ev_function_call(call_id, "update_plan", &invalid_args),
         ev_completed("resp-1"),
     ]);
@@ -260,7 +220,7 @@ async fn update_plan_tool_rejects_malformed_payload() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "malformed plan payload"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -295,22 +255,13 @@ async fn update_plan_tool_rejects_malformed_payload() -> anyhow::Result<()> {
         "did not expect PlanUpdate event for malformed payload"
     );
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert!(!requests.is_empty(), "expected at least one POST request");
-
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect::<Vec<_>>();
-
-    let body_with_tool_output = find_request_with_function_call_output(&request_bodies)
-        .expect("function_call_output item not found in requests");
-    let output_item = function_call_output(body_with_tool_output).expect("tool output item");
+    let req = second_mock.single_request();
+    let output_item = req.function_call_output(call_id);
     assert_eq!(
         output_item.get("call_id").and_then(Value::as_str),
         Some(call_id)
     );
-    let output_text = extract_output_text(output_item).expect("output text present");
+    let output_text = extract_output_text(&output_item).expect("output text present");
     assert!(
         output_text.contains("failed to parse function arguments"),
         "expected parse error message in output text, got {output_text:?}"
@@ -337,7 +288,7 @@ async fn apply_patch_tool_executes_and_emits_patch_events() -> anyhow::Result<()
     let server = start_mock_server().await;
 
     let mut builder = test_codex().with_config(|config| {
-        config.include_apply_patch_tool = true;
+        config.features.enable(Feature::ApplyPatchFreeform);
     });
     let TestCodex {
         codex,
@@ -346,18 +297,19 @@ async fn apply_patch_tool_executes_and_emits_patch_events() -> anyhow::Result<()
         ..
     } = builder.build(&server).await?;
 
+    let file_name = "notes.txt";
+    let file_path = cwd.path().join(file_name);
     let call_id = "apply-patch-call";
-    let patch_content = r#"*** Begin Patch
-*** Add File: notes.txt
+    let patch_content = format!(
+        r#"*** Begin Patch
+*** Add File: {file_name}
 +Tool harness apply patch
-*** End Patch"#;
+*** End Patch"#
+    );
 
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
-        ev_apply_patch_function_call(call_id, patch_content),
+        ev_response_created("resp-1"),
+        ev_apply_patch_function_call(call_id, &patch_content),
         ev_completed("resp-1"),
     ]);
     responses::mount_sse_once_match(&server, any(), first_response).await;
@@ -366,7 +318,7 @@ async fn apply_patch_tool_executes_and_emits_patch_events() -> anyhow::Result<()
         ev_assistant_message("msg-1", "patch complete"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -406,56 +358,31 @@ async fn apply_patch_tool_executes_and_emits_patch_events() -> anyhow::Result<()
     assert!(saw_patch_begin, "expected PatchApplyBegin event");
     let patch_end_success =
         patch_end_success.expect("expected PatchApplyEnd event to capture success flag");
+    assert!(patch_end_success);
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert!(!requests.is_empty(), "expected at least one POST request");
-
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect::<Vec<_>>();
-
-    let body_with_tool_output = find_request_with_function_call_output(&request_bodies)
-        .expect("function_call_output item not found in requests");
-    let output_item = function_call_output(body_with_tool_output).expect("tool output item");
+    let req = second_mock.single_request();
+    let output_item = req.function_call_output(call_id);
     assert_eq!(
         output_item.get("call_id").and_then(Value::as_str),
         Some(call_id)
     );
-    let output_text = extract_output_text(output_item).expect("output text present");
+    let output_text = extract_output_text(&output_item).expect("output text present");
 
-    if let Ok(exec_output) = serde_json::from_str::<Value>(output_text) {
-        let exit_code = exec_output["metadata"]["exit_code"]
-            .as_i64()
-            .expect("exit_code present");
-        let summary = exec_output["output"].as_str().expect("output field");
-        assert_eq!(
-            exit_code, 0,
-            "expected apply_patch exit_code=0, got {exit_code}, summary: {summary:?}"
-        );
-        assert!(
-            patch_end_success,
-            "expected PatchApplyEnd success flag, summary: {summary:?}"
-        );
-        assert!(
-            summary.contains("Success."),
-            "expected apply_patch summary to note success, got {summary:?}"
-        );
+    let expected_pattern = format!(
+        r"(?s)^Exit code: 0
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+Success. Updated the following files:
+A {file_name}
+?$"
+    );
+    assert_regex_match(&expected_pattern, output_text);
 
-        let patched_path = cwd.path().join("notes.txt");
-        let contents = std::fs::read_to_string(&patched_path)
-            .unwrap_or_else(|e| panic!("failed reading {}: {e}", patched_path.display()));
-        assert_eq!(contents, "Tool harness apply patch\n");
-    } else {
-        assert!(
-            output_text.contains("codex-run-as-apply-patch"),
-            "expected apply_patch failure message to mention codex-run-as-apply-patch, got {output_text:?}"
-        );
-        assert!(
-            !patch_end_success,
-            "expected PatchApplyEnd to report success=false when apply_patch invocation fails"
-        );
-    }
+    let updated_contents = fs::read_to_string(file_path)?;
+    assert_eq!(
+        updated_contents, "Tool harness apply patch\n",
+        "expected updated file content"
+    );
 
     Ok(())
 }
@@ -467,7 +394,7 @@ async fn apply_patch_reports_parse_diagnostics() -> anyhow::Result<()> {
     let server = start_mock_server().await;
 
     let mut builder = test_codex().with_config(|config| {
-        config.include_apply_patch_tool = true;
+        config.features.enable(Feature::ApplyPatchFreeform);
     });
     let TestCodex {
         codex,
@@ -482,10 +409,7 @@ async fn apply_patch_reports_parse_diagnostics() -> anyhow::Result<()> {
 *** End Patch";
 
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
+        ev_response_created("resp-1"),
         ev_apply_patch_function_call(call_id, patch_content),
         ev_completed("resp-1"),
     ]);
@@ -495,7 +419,7 @@ async fn apply_patch_reports_parse_diagnostics() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "failed"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let second_mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -516,22 +440,13 @@ async fn apply_patch_reports_parse_diagnostics() -> anyhow::Result<()> {
 
     wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert!(!requests.is_empty(), "expected at least one POST request");
-
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect::<Vec<_>>();
-
-    let body_with_tool_output = find_request_with_function_call_output(&request_bodies)
-        .expect("function_call_output item not found in requests");
-    let output_item = function_call_output(body_with_tool_output).expect("tool output item");
+    let req = second_mock.single_request();
+    let output_item = req.function_call_output(call_id);
     assert_eq!(
         output_item.get("call_id").and_then(Value::as_str),
         Some(call_id)
     );
-    let output_text = extract_output_text(output_item).expect("output text present");
+    let output_text = extract_output_text(&output_item).expect("output text present");
 
     assert!(
         output_text.contains("apply_patch verification failed"),

codex-rs/core/tests/suite/tool_parallelism.rs

@@ -63,8 +63,9 @@ async fn build_codex_with_test_tool(server: &wiremock::MockServer) -> anyhow::Re
 }
 
 fn assert_parallel_duration(actual: Duration) {
+    // Allow headroom for runtime overhead while still differentiating from serial execution.
     assert!(
-        actual < Duration::from_millis(500),
+        actual < Duration::from_millis(750),
         "expected parallel execution to finish quickly, got {actual:?}"
     );
 }
@@ -83,6 +84,16 @@ async fn read_file_tools_run_in_parallel() -> anyhow::Result<()> {
     let server = start_mock_server().await;
     let test = build_codex_with_test_tool(&server).await?;
 
+    let warmup_args = json!({
+        "sleep_after_ms": 10,
+        "barrier": {
+            "id": "parallel-test-sync-warmup",
+            "participants": 2,
+            "timeout_ms": 1_000,
+        }
+    })
+    .to_string();
+
     let parallel_args = json!({
         "sleep_after_ms": 300,
         "barrier": {
@@ -93,6 +104,17 @@ async fn read_file_tools_run_in_parallel() -> anyhow::Result<()> {
     })
     .to_string();
 
+    let warmup_first = sse(vec![
+        json!({"type": "response.created", "response": {"id": "resp-warm-1"}}),
+        ev_function_call("warm-call-1", "test_sync_tool", &warmup_args),
+        ev_function_call("warm-call-2", "test_sync_tool", &warmup_args),
+        ev_completed("resp-warm-1"),
+    ]);
+    let warmup_second = sse(vec![
+        ev_assistant_message("warm-msg-1", "warmup complete"),
+        ev_completed("resp-warm-2"),
+    ]);
+
     let first_response = sse(vec![
         json!({"type": "response.created", "response": {"id": "resp-1"}}),
         ev_function_call("call-1", "test_sync_tool", &parallel_args),
@@ -103,7 +125,13 @@ async fn read_file_tools_run_in_parallel() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    mount_sse_sequence(&server, vec![first_response, second_response]).await;
+    mount_sse_sequence(
+        &server,
+        vec![warmup_first, warmup_second, first_response, second_response],
+    )
+    .await;
+
+    run_turn(&test, "warm up parallel tool").await?;
 
     let duration = run_turn_and_measure(&test, "exercise sync tool").await?;
     assert_parallel_duration(duration);

codex-rs/core/tests/suite/tools.rs

@@ -2,6 +2,7 @@
 #![allow(clippy::unwrap_used, clippy::expect_used)]
 
 use anyhow::Result;
+use codex_core::features::Feature;
 use codex_core::model_family::find_family_for_model;
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::EventMsg;
@@ -9,10 +10,13 @@ use codex_core::protocol::InputItem;
 use codex_core::protocol::Op;
 use codex_core::protocol::SandboxPolicy;
 use codex_protocol::config_types::ReasoningSummary;
+use core_test_support::assert_regex_match;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_custom_tool_call;
 use core_test_support::responses::ev_function_call;
+use core_test_support::responses::ev_response_created;
+use core_test_support::responses::mount_sse_once;
 use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
@@ -20,9 +24,9 @@ use core_test_support::skip_if_no_network;
 use core_test_support::test_codex::TestCodex;
 use core_test_support::test_codex::test_codex;
 use core_test_support::wait_for_event;
+use regex_lite::Regex;
 use serde_json::Value;
 use serde_json::json;
-use wiremock::Request;
 
 async fn submit_turn(
     test: &TestCodex,
@@ -55,27 +59,6 @@ async fn submit_turn(
     Ok(())
 }
 
-fn request_bodies(requests: &[Request]) -> Result<Vec<Value>> {
-    requests
-        .iter()
-        .map(|req| Ok(serde_json::from_slice::<Value>(&req.body)?))
-        .collect()
-}
-
-fn collect_output_items<'a>(bodies: &'a [Value], ty: &str) -> Vec<&'a Value> {
-    let mut out = Vec::new();
-    for body in bodies {
-        if let Some(items) = body.get("input").and_then(Value::as_array) {
-            for item in items {
-                if item.get("type").and_then(Value::as_str) == Some(ty) {
-                    out.push(item);
-                }
-            }
-        }
-    }
-    out
-}
-
 fn tool_names(body: &Value) -> Vec<String> {
     body.get("tools")
         .and_then(Value::as_array)
@@ -104,18 +87,23 @@ async fn custom_tool_unknown_returns_custom_output_error() -> Result<()> {
     let call_id = "custom-unsupported";
     let tool_name = "unsupported_tool";
 
-    let responses = vec![
+    mount_sse_once(
+        &server,
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_custom_tool_call(call_id, tool_name, "\"payload\""),
             ev_completed("resp-1"),
         ]),
+    )
+    .await;
+    let mock = mount_sse_once(
+        &server,
         sse(vec![
             ev_assistant_message("msg-1", "done"),
             ev_completed("resp-2"),
         ]),
-    ];
-    mount_sse_sequence(&server, responses).await;
+    )
+    .await;
 
     submit_turn(
         &test,
@@ -125,13 +113,7 @@ async fn custom_tool_unknown_returns_custom_output_error() -> Result<()> {
     )
     .await?;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    let bodies = request_bodies(&requests)?;
-    let custom_items = collect_output_items(&bodies, "custom_tool_call_output");
-    assert_eq!(custom_items.len(), 1, "expected single custom tool output");
-    let item = custom_items[0];
-    assert_eq!(item.get("call_id").and_then(Value::as_str), Some(call_id));
-
+    let item = mock.single_request().custom_tool_call_output(call_id);
     let output = item
         .get("output")
         .and_then(Value::as_str)
@@ -167,9 +149,10 @@ async fn shell_escalated_permissions_rejected_then_ok() -> Result<()> {
         "timeout_ms": 1_000,
     });
 
-    let responses = vec![
+    mount_sse_once(
+        &server,
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_function_call(
                 call_id_blocked,
                 "shell",
@@ -177,8 +160,12 @@ async fn shell_escalated_permissions_rejected_then_ok() -> Result<()> {
             ),
             ev_completed("resp-1"),
         ]),
+    )
+    .await;
+    let second_mock = mount_sse_once(
+        &server,
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-2"}}),
+            ev_response_created("resp-2"),
             ev_function_call(
                 call_id_success,
                 "shell",
@@ -186,12 +173,16 @@ async fn shell_escalated_permissions_rejected_then_ok() -> Result<()> {
             ),
             ev_completed("resp-2"),
         ]),
+    )
+    .await;
+    let third_mock = mount_sse_once(
+        &server,
         sse(vec![
             ev_assistant_message("msg-1", "done"),
             ev_completed("resp-3"),
         ]),
-    ];
-    mount_sse_sequence(&server, responses).await;
+    )
+    .await;
 
     submit_turn(
         &test,
@@ -201,46 +192,23 @@ async fn shell_escalated_permissions_rejected_then_ok() -> Result<()> {
     )
     .await?;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    let bodies = request_bodies(&requests)?;
-    let function_outputs = collect_output_items(&bodies, "function_call_output");
-    for item in &function_outputs {
-        let call_id = item
-            .get("call_id")
-            .and_then(Value::as_str)
-            .unwrap_or_default();
-        assert!(
-            call_id == call_id_blocked || call_id == call_id_success,
-            "unexpected call id {call_id}"
-        );
-    }
-
     let policy = AskForApproval::Never;
     let expected_message = format!(
         "approval policy is {policy:?}; reject command — you should not ask for escalated permissions if the approval policy is {policy:?}"
     );
 
-    let blocked_outputs: Vec<&Value> = function_outputs
-        .iter()
-        .filter(|item| item.get("call_id").and_then(Value::as_str) == Some(call_id_blocked))
-        .copied()
-        .collect();
-    assert!(
-        !blocked_outputs.is_empty(),
-        "expected at least one rejection output for {call_id_blocked}"
+    let blocked_item = second_mock
+        .single_request()
+        .function_call_output(call_id_blocked);
+    assert_eq!(
+        blocked_item.get("output").and_then(Value::as_str),
+        Some(expected_message.as_str()),
+        "unexpected rejection message"
     );
-    for item in blocked_outputs {
-        assert_eq!(
-            item.get("output").and_then(Value::as_str),
-            Some(expected_message.as_str()),
-            "unexpected rejection message"
-        );
-    }
 
-    let success_item = function_outputs
-        .iter()
-        .find(|item| item.get("call_id").and_then(Value::as_str) == Some(call_id_success))
-        .expect("success output present");
+    let success_item = third_mock
+        .single_request()
+        .function_call_output(call_id_success);
     let output_json: Value = serde_json::from_str(
         success_item
             .get("output")
@@ -253,10 +221,8 @@ async fn shell_escalated_permissions_rejected_then_ok() -> Result<()> {
         "expected exit code 0 after rerunning without escalation",
     );
     let stdout = output_json["output"].as_str().unwrap_or_default();
-    assert!(
-        stdout.contains("shell ok"),
-        "expected stdout to include command output, got {stdout:?}"
-    );
+    let stdout_pattern = r"(?s)^shell ok\n?$";
+    assert_regex_match(stdout_pattern, stdout);
 
     Ok(())
 }
@@ -281,18 +247,23 @@ async fn local_shell_missing_ids_maps_to_function_output_error() -> Result<()> {
         }
     });
 
-    let responses = vec![
+    mount_sse_once(
+        &server,
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             local_shell_event,
             ev_completed("resp-1"),
         ]),
+    )
+    .await;
+    let second_mock = mount_sse_once(
+        &server,
         sse(vec![
             ev_assistant_message("msg-1", "done"),
             ev_completed("resp-2"),
         ]),
-    ];
-    mount_sse_sequence(&server, responses).await;
+    )
+    .await;
 
     submit_turn(
         &test,
@@ -302,15 +273,7 @@ async fn local_shell_missing_ids_maps_to_function_output_error() -> Result<()> {
     )
     .await?;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    let bodies = request_bodies(&requests)?;
-    let function_outputs = collect_output_items(&bodies, "function_call_output");
-    assert_eq!(
-        function_outputs.len(),
-        1,
-        "expected a single function output"
-    );
-    let item = function_outputs[0];
+    let item = second_mock.single_request().function_call_output("");
     assert_eq!(item.get("call_id").and_then(Value::as_str), Some(""));
     assert_eq!(
         item.get("output").and_then(Value::as_str),
@@ -324,14 +287,18 @@ async fn collect_tools(use_unified_exec: bool) -> Result<Vec<String>> {
     let server = start_mock_server().await;
 
     let responses = vec![sse(vec![
-        json!({"type": "response.created", "response": {"id": "resp-1"}}),
+        ev_response_created("resp-1"),
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-1"),
     ])];
-    mount_sse_sequence(&server, responses).await;
+    let mock = mount_sse_sequence(&server, responses).await;
 
     let mut builder = test_codex().with_config(move |config| {
-        config.use_experimental_unified_exec_tool = use_unified_exec;
+        if use_unified_exec {
+            config.features.enable(Feature::UnifiedExec);
+        } else {
+            config.features.disable(Feature::UnifiedExec);
+        }
     });
     let test = builder.build(&server).await?;
 
@@ -343,15 +310,8 @@ async fn collect_tools(use_unified_exec: bool) -> Result<Vec<String>> {
     )
     .await?;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert_eq!(
-        requests.len(),
-        1,
-        "expected a single request for tools collection"
-    );
-    let bodies = request_bodies(&requests)?;
-    let first_body = bodies.first().expect("request body present");
-    Ok(tool_names(first_body))
+    let first_body = mock.single_request().body_json();
+    Ok(tool_names(&first_body))
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
@@ -391,18 +351,23 @@ async fn shell_timeout_includes_timeout_prefix_and_metadata() -> Result<()> {
         "timeout_ms": timeout_ms,
     });
 
-    let responses = vec![
+    mount_sse_once(
+        &server,
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
             ev_completed("resp-1"),
         ]),
+    )
+    .await;
+    let second_mock = mount_sse_once(
+        &server,
         sse(vec![
             ev_assistant_message("msg-1", "done"),
             ev_completed("resp-2"),
         ]),
-    ];
-    mount_sse_sequence(&server, responses).await;
+    )
+    .await;
 
     submit_turn(
         &test,
@@ -412,13 +377,7 @@ async fn shell_timeout_includes_timeout_prefix_and_metadata() -> Result<()> {
     )
     .await?;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    let bodies = request_bodies(&requests)?;
-    let function_outputs = collect_output_items(&bodies, "function_call_output");
-    let timeout_item = function_outputs
-        .iter()
-        .find(|item| item.get("call_id").and_then(Value::as_str) == Some(call_id))
-        .expect("timeout output present");
+    let timeout_item = second_mock.single_request().function_call_output(call_id);
 
     let output_str = timeout_item
         .get("output")
@@ -437,114 +396,18 @@ async fn shell_timeout_includes_timeout_prefix_and_metadata() -> Result<()> {
 
         let stdout = output_json["output"].as_str().unwrap_or_default();
         assert!(
-            stdout.contains("command timed out after "),
-            "expected timeout prefix, got {stdout:?}"
-        );
-        let third_line = stdout.lines().nth(2).unwrap_or_default();
-        let duration_ms = third_line
-            .strip_prefix("command timed out after ")
-            .and_then(|line| line.strip_suffix(" milliseconds"))
-            .and_then(|value| value.parse::<u64>().ok())
-            .unwrap_or_default();
-        assert!(
-            duration_ms >= timeout_ms,
-            "expected duration >= configured timeout, got {duration_ms} (timeout {timeout_ms})"
+            stdout.contains("command timed out"),
+            "timeout output missing `command timed out`: {stdout}"
         );
     } else {
         // Fallback: accept the signal classification path to deflake the test.
-        assert!(
-            output_str.contains("execution error"),
-            "unexpected non-JSON output: {output_str:?}"
-        );
-        assert!(
-            output_str.contains("Signal(") || output_str.to_lowercase().contains("signal"),
-            "expected signal classification in error output, got {output_str:?}"
-        );
+        let signal_pattern = r"(?is)^execution error:.*signal.*$";
+        assert_regex_match(signal_pattern, output_str);
     }
 
     Ok(())
 }
 
-#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn shell_sandbox_denied_truncates_error_output() -> Result<()> {
-    skip_if_no_network!(Ok(()));
-
-    let server = start_mock_server().await;
-    let mut builder = test_codex();
-    let test = builder.build(&server).await?;
-
-    let call_id = "shell-denied";
-    let long_line = "this is a long stderr line that should trigger truncation 0123456789abcdefghijklmnopqrstuvwxyz";
-    let script = format!(
-        "for i in $(seq 1 500); do >&2 echo '{long_line}'; done; cat <<'EOF' > denied.txt\ncontent\nEOF",
-    );
-    let args = json!({
-        "command": ["/bin/sh", "-c", script],
-        "timeout_ms": 1_000,
-    });
-
-    let responses = vec![
-        sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
-            ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
-            ev_completed("resp-1"),
-        ]),
-        sse(vec![
-            ev_assistant_message("msg-1", "done"),
-            ev_completed("resp-2"),
-        ]),
-    ];
-    mount_sse_sequence(&server, responses).await;
-
-    submit_turn(
-        &test,
-        "attempt to write in read-only sandbox",
-        AskForApproval::Never,
-        SandboxPolicy::ReadOnly,
-    )
-    .await?;
-
-    let requests = server.received_requests().await.expect("recorded requests");
-    let bodies = request_bodies(&requests)?;
-    let function_outputs = collect_output_items(&bodies, "function_call_output");
-    let denied_item = function_outputs
-        .iter()
-        .find(|item| item.get("call_id").and_then(Value::as_str) == Some(call_id))
-        .expect("denied output present");
-
-    let output = denied_item
-        .get("output")
-        .and_then(Value::as_str)
-        .expect("denied output string");
-
-    assert!(
-        output.contains("failed in sandbox: "),
-        "expected sandbox error prefix, got {output:?}"
-    );
-    assert!(
-        output.contains("[... omitted"),
-        "expected truncated marker, got {output:?}"
-    );
-    assert!(
-        output.contains(long_line),
-        "expected truncated stderr sample, got {output:?}"
-    );
-    // Linux distributions may surface sandbox write failures as different errno messages
-    // depending on the underlying mechanism (e.g., EPERM, EACCES, or EROFS). Accept a
-    // small set of common variants to keep this cross-platform.
-    let denial_markers = [
-        "Operation not permitted", // EPERM
-        "Permission denied",       // EACCES
-        "Read-only file system",   // EROFS
-    ];
-    assert!(
-        denial_markers.iter().any(|m| output.contains(m)),
-        "expected sandbox denial message, got {output:?}"
-    );
-
-    Ok(())
-}
-
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn shell_spawn_failure_truncates_exec_error() -> Result<()> {
     skip_if_no_network!(Ok(()));
@@ -569,18 +432,23 @@ async fn shell_spawn_failure_truncates_exec_error() -> Result<()> {
         "timeout_ms": 1_000,
     });
 
-    let responses = vec![
+    mount_sse_once(
+        &server,
         sse(vec![
-            json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_function_call(call_id, "shell", &serde_json::to_string(&args)?),
             ev_completed("resp-1"),
         ]),
+    )
+    .await;
+    let second_mock = mount_sse_once(
+        &server,
         sse(vec![
             ev_assistant_message("msg-1", "done"),
             ev_completed("resp-2"),
         ]),
-    ];
-    mount_sse_sequence(&server, responses).await;
+    )
+    .await;
 
     submit_turn(
         &test,
@@ -590,23 +458,29 @@ async fn shell_spawn_failure_truncates_exec_error() -> Result<()> {
     )
     .await?;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    let bodies = request_bodies(&requests)?;
-    let function_outputs = collect_output_items(&bodies, "function_call_output");
-    let failure_item = function_outputs
-        .iter()
-        .find(|item| item.get("call_id").and_then(Value::as_str) == Some(call_id))
-        .expect("spawn failure output present");
+    let failure_item = second_mock.single_request().function_call_output(call_id);
 
     let output = failure_item
         .get("output")
         .and_then(Value::as_str)
         .expect("spawn failure output string");
 
-    assert!(
-        output.contains("execution error:"),
-        "expected execution error prefix, got {output:?}"
-    );
+    let spawn_error_pattern = r#"(?s)^Exit code: -?\d+
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Output:
+execution error: .*$"#;
+    let spawn_truncated_pattern = r#"(?s)^Exit code: -?\d+
+Wall time: [0-9]+(?:\.[0-9]+)? seconds
+Total output lines: \d+
+Output:
+
+execution error: .*$"#;
+    let spawn_error_regex = Regex::new(spawn_error_pattern)?;
+    let spawn_truncated_regex = Regex::new(spawn_truncated_pattern)?;
+    if !spawn_error_regex.is_match(output) && !spawn_truncated_regex.is_match(output) {
+        let fallback_pattern = r"(?s)^execution error: .*$";
+        assert_regex_match(fallback_pattern, output);
+    }
     assert!(output.len() <= 10 * 1024);
 
     Ok(())

codex-rs/core/tests/suite/unified_exec.rs

@@ -3,6 +3,7 @@
 use std::collections::HashMap;
 
 use anyhow::Result;
+use codex_core::features::Feature;
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::EventMsg;
 use codex_core::protocol::InputItem;
@@ -12,6 +13,7 @@ use codex_protocol::config_types::ReasoningSummary;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_function_call;
+use core_test_support::responses::ev_response_created;
 use core_test_support::responses::mount_sse_sequence;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
@@ -41,7 +43,13 @@ fn collect_tool_outputs(bodies: &[Value]) -> Result<HashMap<String, Value>> {
                 if let Some(call_id) = item.get("call_id").and_then(Value::as_str) {
                     let content = extract_output_text(item)
                         .ok_or_else(|| anyhow::anyhow!("missing tool output content"))?;
-                    let parsed: Value = serde_json::from_str(content)?;
+                    let trimmed = content.trim();
+                    if trimmed.is_empty() {
+                        continue;
+                    }
+                    let parsed: Value = serde_json::from_str(trimmed).map_err(|err| {
+                        anyhow::anyhow!("failed to parse tool output content {trimmed:?}: {err}")
+                    })?;
                     outputs.insert(call_id.to_string(), parsed);
                 }
             }
@@ -58,7 +66,7 @@ async fn unified_exec_reuses_session_via_stdin() -> Result<()> {
     let server = start_mock_server().await;
 
     let mut builder = test_codex().with_config(|config| {
-        config.use_experimental_unified_exec_tool = true;
+        config.features.enable(Feature::UnifiedExec);
     });
     let TestCodex {
         codex,
@@ -82,7 +90,7 @@ async fn unified_exec_reuses_session_via_stdin() -> Result<()> {
 
     let responses = vec![
         sse(vec![
-            serde_json::json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_function_call(
                 first_call_id,
                 "unified_exec",
@@ -91,7 +99,7 @@ async fn unified_exec_reuses_session_via_stdin() -> Result<()> {
             ev_completed("resp-1"),
         ]),
         sse(vec![
-            serde_json::json!({"type": "response.created", "response": {"id": "resp-2"}}),
+            ev_response_created("resp-2"),
             ev_function_call(
                 second_call_id,
                 "unified_exec",
@@ -167,7 +175,7 @@ async fn unified_exec_reuses_session_via_stdin() -> Result<()> {
 }
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-async fn unified_exec_timeout_and_followup_poll() -> Result<()> {
+async fn unified_exec_streams_after_lagged_output() -> Result<()> {
     skip_if_no_network!(Ok(()));
     skip_if_sandbox!(Ok(()));
 
@@ -175,6 +183,132 @@ async fn unified_exec_timeout_and_followup_poll() -> Result<()> {
 
     let mut builder = test_codex().with_config(|config| {
         config.use_experimental_unified_exec_tool = true;
+        config.features.enable(Feature::UnifiedExec);
+    });
+    let TestCodex {
+        codex,
+        cwd,
+        session_configured,
+        ..
+    } = builder.build(&server).await?;
+
+    let script = r#"python3 - <<'PY'
+import sys
+import time
+
+chunk = b'x' * (1 << 20)
+for _ in range(4):
+    sys.stdout.buffer.write(chunk)
+    sys.stdout.flush()
+
+time.sleep(0.2)
+for _ in range(5):
+    sys.stdout.write("TAIL-MARKER\n")
+    sys.stdout.flush()
+    time.sleep(0.05)
+
+time.sleep(0.2)
+PY
+"#;
+
+    let first_call_id = "uexec-lag-start";
+    let first_args = serde_json::json!({
+        "input": ["/bin/sh", "-c", script],
+        "timeout_ms": 25,
+    });
+
+    let second_call_id = "uexec-lag-poll";
+    let second_args = serde_json::json!({
+        "input": Vec::<String>::new(),
+        "session_id": "0",
+        "timeout_ms": 2_000,
+    });
+
+    let responses = vec![
+        sse(vec![
+            ev_response_created("resp-1"),
+            ev_function_call(
+                first_call_id,
+                "unified_exec",
+                &serde_json::to_string(&first_args)?,
+            ),
+            ev_completed("resp-1"),
+        ]),
+        sse(vec![
+            ev_response_created("resp-2"),
+            ev_function_call(
+                second_call_id,
+                "unified_exec",
+                &serde_json::to_string(&second_args)?,
+            ),
+            ev_completed("resp-2"),
+        ]),
+        sse(vec![
+            ev_assistant_message("msg-1", "lag handled"),
+            ev_completed("resp-3"),
+        ]),
+    ];
+    mount_sse_sequence(&server, responses).await;
+
+    let session_model = session_configured.model.clone();
+
+    codex
+        .submit(Op::UserTurn {
+            items: vec![InputItem::Text {
+                text: "exercise lag handling".into(),
+            }],
+            final_output_json_schema: None,
+            cwd: cwd.path().to_path_buf(),
+            approval_policy: AskForApproval::Never,
+            sandbox_policy: SandboxPolicy::DangerFullAccess,
+            model: session_model,
+            effort: None,
+            summary: ReasoningSummary::Auto,
+        })
+        .await?;
+
+    wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
+
+    let requests = server.received_requests().await.expect("recorded requests");
+    assert!(!requests.is_empty(), "expected at least one POST request");
+
+    let bodies = requests
+        .iter()
+        .map(|req| req.body_json::<Value>().expect("request json"))
+        .collect::<Vec<_>>();
+
+    let outputs = collect_tool_outputs(&bodies)?;
+
+    let start_output = outputs
+        .get(first_call_id)
+        .expect("missing initial unified_exec output");
+    let session_id = start_output["session_id"].as_str().unwrap_or_default();
+    assert!(
+        !session_id.is_empty(),
+        "expected session id from initial unified_exec response"
+    );
+
+    let poll_output = outputs
+        .get(second_call_id)
+        .expect("missing poll unified_exec output");
+    let poll_text = poll_output["output"].as_str().unwrap_or_default();
+    assert!(
+        poll_text.contains("TAIL-MARKER"),
+        "expected poll output to contain tail marker, got {poll_text:?}"
+    );
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn unified_exec_timeout_and_followup_poll() -> Result<()> {
+    skip_if_no_network!(Ok(()));
+    skip_if_sandbox!(Ok(()));
+
+    let server = start_mock_server().await;
+
+    let mut builder = test_codex().with_config(|config| {
+        config.features.enable(Feature::UnifiedExec);
     });
     let TestCodex {
         codex,
@@ -198,7 +332,7 @@ async fn unified_exec_timeout_and_followup_poll() -> Result<()> {
 
     let responses = vec![
         sse(vec![
-            serde_json::json!({"type": "response.created", "response": {"id": "resp-1"}}),
+            ev_response_created("resp-1"),
             ev_function_call(
                 first_call_id,
                 "unified_exec",
@@ -207,7 +341,7 @@ async fn unified_exec_timeout_and_followup_poll() -> Result<()> {
             ev_completed("resp-1"),
         ]),
         sse(vec![
-            serde_json::json!({"type": "response.created", "response": {"id": "resp-2"}}),
+            ev_response_created("resp-2"),
             ev_function_call(
                 second_call_id,
                 "unified_exec",

codex-rs/core/tests/suite/user_notification.rs

@@ -5,6 +5,7 @@ use std::os::unix::fs::PermissionsExt;
 use codex_core::protocol::EventMsg;
 use codex_core::protocol::InputItem;
 use codex_core::protocol::Op;
+use core_test_support::fs_wait;
 use core_test_support::responses;
 use core_test_support::skip_if_no_network;
 use core_test_support::test_codex::TestCodex;
@@ -17,8 +18,7 @@ use responses::ev_assistant_message;
 use responses::ev_completed;
 use responses::sse;
 use responses::start_mock_server;
-use tokio::time::Duration;
-use tokio::time::sleep;
+use std::time::Duration;
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn summarize_context_three_requests_and_instructions() -> anyhow::Result<()> {
@@ -60,14 +60,7 @@ echo -n "${@: -1}" > $(dirname "${0}")/notify.txt"#,
     wait_for_event(&codex, |ev| matches!(ev, EventMsg::TaskComplete(_))).await;
 
     // We fork the notify script, so we need to wait for it to write to the file.
-    for _ in 0..100u32 {
-        if notify_file.exists() {
-            break;
-        }
-        sleep(Duration::from_millis(100)).await;
-    }
-
-    assert!(notify_file.exists());
+    fs_wait::wait_for_path_exists(&notify_file, Duration::from_secs(5)).await?;
 
     Ok(())
 }

codex-rs/core/tests/suite/view_image.rs

@@ -12,6 +12,7 @@ use core_test_support::responses;
 use core_test_support::responses::ev_assistant_message;
 use core_test_support::responses::ev_completed;
 use core_test_support::responses::ev_function_call;
+use core_test_support::responses::ev_response_created;
 use core_test_support::responses::sse;
 use core_test_support::responses::start_mock_server;
 use core_test_support::skip_if_no_network;
@@ -21,16 +22,6 @@ use core_test_support::wait_for_event;
 use serde_json::Value;
 use wiremock::matchers::any;
 
-fn function_call_output(body: &Value) -> Option<&Value> {
-    body.get("input")
-        .and_then(Value::as_array)
-        .and_then(|items| {
-            items.iter().find(|item| {
-                item.get("type").and_then(Value::as_str) == Some("function_call_output")
-            })
-        })
-}
-
 fn find_image_message(body: &Value) -> Option<&Value> {
     body.get("input")
         .and_then(Value::as_array)
@@ -58,12 +49,6 @@ fn extract_output_text(item: &Value) -> Option<&str> {
     })
 }
 
-fn find_request_with_function_call_output(requests: &[Value]) -> Option<&Value> {
-    requests
-        .iter()
-        .find(|body| function_call_output(body).is_some())
-}
-
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn view_image_tool_attaches_local_image() -> anyhow::Result<()> {
     skip_if_no_network!(Ok(()));
@@ -89,10 +74,7 @@ async fn view_image_tool_attaches_local_image() -> anyhow::Result<()> {
     let arguments = serde_json::json!({ "path": rel_path }).to_string();
 
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
+        ev_response_created("resp-1"),
         ev_function_call(call_id, "view_image", &arguments),
         ev_completed("resp-1"),
     ]);
@@ -102,7 +84,7 @@ async fn view_image_tool_attaches_local_image() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -139,25 +121,14 @@ async fn view_image_tool_attaches_local_image() -> anyhow::Result<()> {
     assert_eq!(tool_event.call_id, call_id);
     assert_eq!(tool_event.path, abs_path);
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert!(
-        requests.len() >= 2,
-        "expected at least two POST requests, got {}",
-        requests.len()
-    );
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect::<Vec<_>>();
+    let body = mock.single_request().body_json();
+    let output_item = mock.single_request().function_call_output(call_id);
 
-    let body_with_tool_output = find_request_with_function_call_output(&request_bodies)
-        .expect("function_call_output item not found in requests");
-    let output_item = function_call_output(body_with_tool_output).expect("tool output item");
-    let output_text = extract_output_text(output_item).expect("output text present");
+    let output_text = extract_output_text(&output_item).expect("output text present");
     assert_eq!(output_text, "attached local image path");
 
-    let image_message = find_image_message(body_with_tool_output)
-        .expect("pending input image message not included in request");
+    let image_message =
+        find_image_message(&body).expect("pending input image message not included in request");
     let image_url = image_message
         .get("content")
         .and_then(Value::as_array)
@@ -202,10 +173,7 @@ async fn view_image_tool_errors_when_path_is_directory() -> anyhow::Result<()> {
     let arguments = serde_json::json!({ "path": rel_path }).to_string();
 
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
+        ev_response_created("resp-1"),
         ev_function_call(call_id, "view_image", &arguments),
         ev_completed("resp-1"),
     ]);
@@ -215,7 +183,7 @@ async fn view_image_tool_errors_when_path_is_directory() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -236,26 +204,14 @@ async fn view_image_tool_errors_when_path_is_directory() -> anyhow::Result<()> {
 
     wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert!(
-        requests.len() >= 2,
-        "expected at least two POST requests, got {}",
-        requests.len()
-    );
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect::<Vec<_>>();
-
-    let body_with_tool_output = find_request_with_function_call_output(&request_bodies)
-        .expect("function_call_output item not found in requests");
-    let output_item = function_call_output(body_with_tool_output).expect("tool output item");
-    let output_text = extract_output_text(output_item).expect("output text present");
+    let body_with_tool_output = mock.single_request().body_json();
+    let output_item = mock.single_request().function_call_output(call_id);
+    let output_text = extract_output_text(&output_item).expect("output text present");
     let expected_message = format!("image path `{}` is not a file", abs_path.display());
     assert_eq!(output_text, expected_message);
 
     assert!(
-        find_image_message(body_with_tool_output).is_none(),
+        find_image_message(&body_with_tool_output).is_none(),
         "directory path should not produce an input_image message"
     );
 
@@ -282,10 +238,7 @@ async fn view_image_tool_errors_when_file_missing() -> anyhow::Result<()> {
     let arguments = serde_json::json!({ "path": rel_path }).to_string();
 
     let first_response = sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp-1"}
-        }),
+        ev_response_created("resp-1"),
         ev_function_call(call_id, "view_image", &arguments),
         ev_completed("resp-1"),
     ]);
@@ -295,7 +248,7 @@ async fn view_image_tool_errors_when_file_missing() -> anyhow::Result<()> {
         ev_assistant_message("msg-1", "done"),
         ev_completed("resp-2"),
     ]);
-    responses::mount_sse_once_match(&server, any(), second_response).await;
+    let mock = responses::mount_sse_once_match(&server, any(), second_response).await;
 
     let session_model = session_configured.model.clone();
 
@@ -316,21 +269,9 @@ async fn view_image_tool_errors_when_file_missing() -> anyhow::Result<()> {
 
     wait_for_event(&codex, |event| matches!(event, EventMsg::TaskComplete(_))).await;
 
-    let requests = server.received_requests().await.expect("recorded requests");
-    assert!(
-        requests.len() >= 2,
-        "expected at least two POST requests, got {}",
-        requests.len()
-    );
-    let request_bodies = requests
-        .iter()
-        .map(|req| req.body_json::<Value>().expect("request json"))
-        .collect::<Vec<_>>();
-
-    let body_with_tool_output = find_request_with_function_call_output(&request_bodies)
-        .expect("function_call_output item not found in requests");
-    let output_item = function_call_output(body_with_tool_output).expect("tool output item");
-    let output_text = extract_output_text(output_item).expect("output text present");
+    let body_with_tool_output = mock.single_request().body_json();
+    let output_item = mock.single_request().function_call_output(call_id);
+    let output_text = extract_output_text(&output_item).expect("output text present");
     let expected_prefix = format!("unable to locate image at `{}`:", abs_path.display());
     assert!(
         output_text.starts_with(&expected_prefix),
@@ -338,7 +279,7 @@ async fn view_image_tool_errors_when_file_missing() -> anyhow::Result<()> {
     );
 
     assert!(
-        find_image_message(body_with_tool_output).is_none(),
+        find_image_message(&body_with_tool_output).is_none(),
         "missing file should not produce an input_image message"
     );
 

codex-rs/docs/codex_mcp_interface.md

@@ -1,6 +1,6 @@
-# Codex MCP Interface [experimental]
+# Codex MCP Server Interface [experimental]
 
-This document describes Codex’s experimental MCP interface: a JSON‑RPC API that runs over the Model Context Protocol (MCP) transport to control a local Codex engine.
+This document describes Codex’s experimental MCP server interface: a JSON‑RPC API that runs over the Model Context Protocol (MCP) transport to control a local Codex engine.
 
 - Status: experimental and subject to change without notice
 - Server binary: `codex mcp-server` (or `codex-mcp-server`)

codex-rs/exec/src/cli.rs

@@ -77,7 +77,7 @@ pub struct Cli {
 
     /// Initial instructions for the agent. If not provided as an argument (or
     /// if `-` is used), instructions are read from stdin.
-    #[arg(value_name = "PROMPT")]
+    #[arg(value_name = "PROMPT", value_hint = clap::ValueHint::Other)]
     pub prompt: Option<String>,
 }
 
@@ -99,7 +99,7 @@ pub struct ResumeArgs {
     pub last: bool,
 
     /// Prompt to send after resuming the session. If `-` is used, read from stdin.
-    #[arg(value_name = "PROMPT")]
+    #[arg(value_name = "PROMPT", value_hint = clap::ValueHint::Other)]
     pub prompt: Option<String>,
 }
 

codex-rs/exec/src/lib.rs

@@ -48,7 +48,7 @@ use codex_core::default_client::set_default_originator;
 use codex_core::find_conversation_path_by_id_str;
 
 pub async fn run_main(cli: Cli, codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()> {
-    if let Err(err) = set_default_originator("codex_exec") {
+    if let Err(err) = set_default_originator("codex_exec".to_string()) {
         tracing::warn!(?err, "Failed to set codex exec originator override {err:?}");
     }
 

codex-rs/exec/tests/suite/auth_env.rs

@@ -1,26 +1,22 @@
 #![allow(clippy::unwrap_used, clippy::expect_used)]
 use core_test_support::responses::ev_completed;
+use core_test_support::responses::mount_sse_once_match;
 use core_test_support::responses::sse;
-use core_test_support::responses::sse_response;
 use core_test_support::responses::start_mock_server;
 use core_test_support::test_codex_exec::test_codex_exec;
-use wiremock::Mock;
 use wiremock::matchers::header;
-use wiremock::matchers::method;
-use wiremock::matchers::path;
 
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn exec_uses_codex_api_key_env_var() -> anyhow::Result<()> {
     let test = test_codex_exec();
     let server = start_mock_server().await;
 
-    Mock::given(method("POST"))
-        .and(path("/v1/responses"))
-        .and(header("Authorization", "Bearer dummy"))
-        .respond_with(sse_response(sse(vec![ev_completed("request_0")])))
-        .expect(1)
-        .mount(&server)
-        .await;
+    mount_sse_once_match(
+        &server,
+        header("Authorization", "Bearer dummy"),
+        sse(vec![ev_completed("request_0")]),
+    )
+    .await;
 
     test.cmd_with_server(&server)
         .arg("--skip-git-repo-check")

codex-rs/exec/tests/suite/mod.rs

@@ -1,6 +1,7 @@
 // Aggregates all former standalone integration tests as modules.
 mod apply_patch;
 mod auth_env;
+mod originator;
 mod output_schema;
 mod resume;
 mod sandbox;

codex-rs/exec/tests/suite/originator.rs

@@ -0,0 +1,52 @@
+#![cfg(not(target_os = "windows"))]
+#![allow(clippy::expect_used, clippy::unwrap_used)]
+
+use core_test_support::responses;
+use core_test_support::test_codex_exec::test_codex_exec;
+use wiremock::matchers::header;
+
+/// Verify that when the server reports an error, `codex-exec` exits with a
+/// non-zero status code so automation can detect failures.
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn send_codex_exec_originator() -> anyhow::Result<()> {
+    let test = test_codex_exec();
+
+    let server = responses::start_mock_server().await;
+    let body = responses::sse(vec![
+        responses::ev_response_created("response_1"),
+        responses::ev_assistant_message("response_1", "Hello, world!"),
+        responses::ev_completed("response_1"),
+    ]);
+    responses::mount_sse_once_match(&server, header("Originator", "codex_exec"), body).await;
+
+    test.cmd_with_server(&server)
+        .arg("--skip-git-repo-check")
+        .arg("tell me something")
+        .assert()
+        .code(0);
+
+    Ok(())
+}
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn supports_originator_override() -> anyhow::Result<()> {
+    let test = test_codex_exec();
+
+    let server = responses::start_mock_server().await;
+    let body = responses::sse(vec![
+        responses::ev_response_created("response_1"),
+        responses::ev_assistant_message("response_1", "Hello, world!"),
+        responses::ev_completed("response_1"),
+    ]);
+    responses::mount_sse_once_match(&server, header("Originator", "codex_exec_override"), body)
+        .await;
+
+    test.cmd_with_server(&server)
+        .env("CODEX_INTERNAL_ORIGINATOR_OVERRIDE", "codex_exec_override")
+        .arg("--skip-git-repo-check")
+        .arg("tell me something")
+        .assert()
+        .code(0);
+
+    Ok(())
+}

codex-rs/exec/tests/suite/output_schema.rs

@@ -24,14 +24,11 @@ async fn exec_includes_output_schema_in_request() -> anyhow::Result<()> {
 
     let server = responses::start_mock_server().await;
     let body = responses::sse(vec![
-        serde_json::json!({
-            "type": "response.created",
-            "response": {"id": "resp1"}
-        }),
+        responses::ev_response_created("resp1"),
         responses::ev_assistant_message("m1", "fixture hello"),
         responses::ev_completed("resp1"),
     ]);
-    responses::mount_sse_once_match(&server, any(), body).await;
+    let response_mock = responses::mount_sse_once_match(&server, any(), body).await;
 
     test.cmd_with_server(&server)
         .arg("--skip-git-repo-check")
@@ -46,12 +43,8 @@ async fn exec_includes_output_schema_in_request() -> anyhow::Result<()> {
         .assert()
         .success();
 
-    let requests = server
-        .received_requests()
-        .await
-        .expect("failed to capture requests");
-    assert_eq!(requests.len(), 1, "expected exactly one request");
-    let payload: Value = serde_json::from_slice(&requests[0].body)?;
+    let request = response_mock.single_request();
+    let payload: Value = request.body_json();
     let text = payload.get("text").expect("request missing text field");
     let format = text
         .get("format")

codex-rs/git-tooling/Cargo.toml

@@ -17,4 +17,5 @@ walkdir = "2"
 workspace = true
 
 [dev-dependencies]
+assert_matches = { workspace = true }
 pretty_assertions = "1.4.1"

codex-rs/git-tooling/src/ghost_commits.rs

@@ -186,6 +186,7 @@ fn default_commit_identity() -> Vec<(OsString, OsString)> {
 mod tests {
     use super::*;
     use crate::operations::run_git_for_stdout;
+    use assert_matches::assert_matches;
     use pretty_assertions::assert_eq;
     use std::process::Command;
 
@@ -348,7 +349,7 @@ mod tests {
         let options = CreateGhostCommitOptions::new(repo)
             .force_include(vec![PathBuf::from("../outside.txt")]);
         let err = create_ghost_commit(&options).unwrap_err();
-        assert!(matches!(err, GitToolingError::PathEscapesRepository { .. }));
+        assert_matches!(err, GitToolingError::PathEscapesRepository { .. });
     }
 
     #[test]
@@ -356,7 +357,7 @@ mod tests {
     fn restore_requires_git_repository() {
         let temp = tempfile::tempdir().expect("tempdir");
         let err = restore_to_commit(temp.path(), "deadbeef").unwrap_err();
-        assert!(matches!(err, GitToolingError::NotAGitRepository { .. }));
+        assert_matches!(err, GitToolingError::NotAGitRepository { .. });
     }
 
     #[test]

codex-rs/ollama/Cargo.toml

@@ -28,3 +28,4 @@ tracing = { workspace = true, features = ["log"] }
 wiremock = { workspace = true }
 
 [dev-dependencies]
+assert_matches = { workspace = true }

codex-rs/ollama/src/parser.rs

@@ -30,19 +30,21 @@ pub(crate) fn pull_events_from_value(value: &JsonValue) -> Vec<PullEvent> {
 
 #[cfg(test)]
 mod tests {
+    use assert_matches::assert_matches;
+
     use super::*;
 
     #[test]
     fn test_pull_events_decoder_status_and_success() {
         let v: JsonValue = serde_json::json!({"status":"verifying"});
         let events = pull_events_from_value(&v);
-        assert!(matches!(events.as_slice(), [PullEvent::Status(s)] if s == "verifying"));
+        assert_matches!(events.as_slice(), [PullEvent::Status(s)] if s == "verifying");
 
         let v2: JsonValue = serde_json::json!({"status":"success"});
         let events2 = pull_events_from_value(&v2);
         assert_eq!(events2.len(), 2);
-        assert!(matches!(events2[0], PullEvent::Status(ref s) if s == "success"));
-        assert!(matches!(events2[1], PullEvent::Success));
+        assert_matches!(events2[0], PullEvent::Status(ref s) if s == "success");
+        assert_matches!(events2[1], PullEvent::Success);
     }
 
     #[test]
@@ -50,33 +52,24 @@ mod tests {
         let v: JsonValue = serde_json::json!({"digest":"sha256:abc","total":100});
         let events = pull_events_from_value(&v);
         assert_eq!(events.len(), 1);
-        match &events[0] {
+        assert_matches!(
+            &events[0],
             PullEvent::ChunkProgress {
                 digest,
                 total,
                 completed,
-            } => {
-                assert_eq!(digest, "sha256:abc");
-                assert_eq!(*total, Some(100));
-                assert_eq!(*completed, None);
-            }
-            _ => panic!("expected ChunkProgress"),
-        }
-
+            } if digest == "sha256:abc" && total == &Some(100) && completed.is_none()
+        );
         let v2: JsonValue = serde_json::json!({"digest":"sha256:def","completed":42});
         let events2 = pull_events_from_value(&v2);
         assert_eq!(events2.len(), 1);
-        match &events2[0] {
+        assert_matches!(
+            &events2[0],
             PullEvent::ChunkProgress {
                 digest,
                 total,
                 completed,
-            } => {
-                assert_eq!(digest, "sha256:def");
-                assert_eq!(*total, None);
-                assert_eq!(*completed, Some(42));
-            }
-            _ => panic!("expected ChunkProgress"),
-        }
+            } if digest == "sha256:def" && total.is_none() && completed == &Some(42)
+        );
     }
 }

codex-rs/otel/src/otel_event_manager.rs

@@ -148,21 +148,15 @@ impl OtelEventManager {
         response
     }
 
-    pub async fn log_sse_event<Next, Fut, E>(
+    pub fn log_sse_event<E>(
         &self,
-        next: Next,
-    ) -> Result<Option<Result<StreamEvent, StreamError<E>>>, Elapsed>
-    where
-        Next: FnOnce() -> Fut,
-        Fut: Future<Output = Result<Option<Result<StreamEvent, StreamError<E>>>, Elapsed>>,
+        response: &Result<Option<Result<StreamEvent, StreamError<E>>>, Elapsed>,
+        duration: Duration,
+    ) where
         E: Display,
     {
-        let start = std::time::Instant::now();
-        let response = next().await;
-        let duration = start.elapsed();
-
         match response {
-            Ok(Some(Ok(ref sse))) => {
+            Ok(Some(Ok(sse))) => {
                 if sse.data.trim() == "[DONE]" {
                     self.sse_event(&sse.event, duration);
                 } else {
@@ -191,7 +185,7 @@ impl OtelEventManager {
                     }
                 }
             }
-            Ok(Some(Err(ref error))) => {
+            Ok(Some(Err(error))) => {
                 self.sse_event_failed(None, duration, error);
             }
             Ok(None) => {}
@@ -199,8 +193,6 @@ impl OtelEventManager {
                 self.sse_event_failed(None, duration, &"idle timeout waiting for SSE");
             }
         }
-
-        response
     }
 
     fn sse_event(&self, kind: &str, duration: Duration) {

codex-rs/protocol/src/protocol.rs

@@ -548,10 +548,15 @@ pub struct TaskStartedEvent {
 
 #[derive(Debug, Clone, Deserialize, Serialize, Default, TS)]
 pub struct TokenUsage {
+    #[ts(type = "number")]
     pub input_tokens: u64,
+    #[ts(type = "number")]
     pub cached_input_tokens: u64,
+    #[ts(type = "number")]
     pub output_tokens: u64,
+    #[ts(type = "number")]
     pub reasoning_output_tokens: u64,
+    #[ts(type = "number")]
     pub total_tokens: u64,
 }
 
@@ -559,6 +564,7 @@ pub struct TokenUsage {
 pub struct TokenUsageInfo {
     pub total_token_usage: TokenUsage,
     pub last_token_usage: TokenUsage,
+    #[ts(type = "number | null")]
     pub model_context_window: Option<u64>,
 }
 
@@ -634,8 +640,10 @@ pub struct RateLimitWindow {
     /// Percentage (0-100) of the window that has been consumed.
     pub used_percent: f64,
     /// Rolling window duration, in minutes.
+    #[ts(type = "number | null")]
     pub window_minutes: Option<u64>,
     /// Seconds until the window resets.
+    #[ts(type = "number | null")]
     pub resets_in_seconds: Option<u64>,
 }
 
@@ -1195,6 +1203,11 @@ pub struct StreamErrorEvent {
     pub message: String,
 }
 
+#[derive(Debug, Clone, Deserialize, Serialize, TS)]
+pub struct StreamInfoEvent {
+    pub message: String,
+}
+
 #[derive(Debug, Clone, Deserialize, Serialize, TS)]
 pub struct PatchApplyBeginEvent {
     /// Identifier so this can be paired with the PatchApplyEnd event.
@@ -1235,6 +1248,30 @@ pub struct GetHistoryEntryResponseEvent {
 pub struct McpListToolsResponseEvent {
     /// Fully qualified tool name -> tool definition.
     pub tools: std::collections::HashMap<String, McpTool>,
+    /// Authentication status for each configured MCP server.
+    pub auth_statuses: std::collections::HashMap<String, McpAuthStatus>,
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, TS)]
+#[serde(rename_all = "snake_case")]
+#[ts(rename_all = "snake_case")]
+pub enum McpAuthStatus {
+    Unsupported,
+    NotLoggedIn,
+    BearerToken,
+    OAuth,
+}
+
+impl fmt::Display for McpAuthStatus {
+    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
+        let text = match self {
+            McpAuthStatus::Unsupported => "Unsupported",
+            McpAuthStatus::NotLoggedIn => "Not logged in",
+            McpAuthStatus::BearerToken => "Bearer token",
+            McpAuthStatus::OAuth => "OAuth",
+        };
+        f.write_str(text)
+    }
 }
 
 /// Response payload for `Op::ListCustomPrompts`.