release

feat: add shell snapshot for shell command (#7786 )
feat: warning for long snapshots (#7870 )
2026-02-06 17:03:42 +00:00 · 2025-12-11 08:07:12 -08:00 · 2025-12-11 13:46:43 +00:00 · 2025-12-11 12:42:47 +00:00 · 2025-12-11 00:39:47 -08:00 · 2025-12-10 21:23:16 -08:00
1135 changed files with 77762 additions and 6194 deletions
--- a/.codespellignore
+++ b/.codespellignore
@@ -1 +1,2 @@
 iTerm
+psuedo
--- a/.github/actions/linux-code-sign/action.yml
+++ b/.github/actions/linux-code-sign/action.yml
@@ -0,0 +1,44 @@
+name: linux-code-sign
+description: Sign Linux artifacts with cosign.
+inputs:
+  target:
+    description: Target triple for the artifacts to sign.
+    required: true
+  artifacts-dir:
+    description: Absolute path to the directory containing built binaries to sign.
+    required: true
+
+runs:
+  using: composite
+  steps:
+    - name: Install cosign
+      uses: sigstore/cosign-installer@v3.7.0
+
+    - name: Cosign Linux artifacts
+      shell: bash
+      env:
+        COSIGN_EXPERIMENTAL: "1"
+        COSIGN_YES: "true"
+        COSIGN_OIDC_CLIENT_ID: "sigstore"
+        COSIGN_OIDC_ISSUER: "https://oauth2.sigstore.dev/auth"
+      run: |
+        set -euo pipefail
+
+        dest="${{ inputs.artifacts-dir }}"
+        if [[ ! -d "$dest" ]]; then
+          echo "Destination $dest does not exist"
+          exit 1
+        fi
+
+        for binary in codex codex-responses-api-proxy; do
+          artifact="${dest}/${binary}"
+          if [[ ! -f "$artifact" ]]; then
+            echo "Binary $artifact not found"
+            exit 1
+          fi
+
+          cosign sign-blob \
+            --yes \
+            --bundle "${artifact}.sigstore" \
+            "$artifact"
+        done
--- a/.github/actions/windows-code-sign/action.yml
+++ b/.github/actions/windows-code-sign/action.yml
@@ -0,0 +1,55 @@
+name: windows-code-sign
+description: Sign Windows binaries with Azure Trusted Signing.
+inputs:
+  target:
+    description: Target triple for the artifacts to sign.
+    required: true
+  client-id:
+    description: Azure Trusted Signing client ID.
+    required: true
+  tenant-id:
+    description: Azure tenant ID for Trusted Signing.
+    required: true
+  subscription-id:
+    description: Azure subscription ID for Trusted Signing.
+    required: true
+  endpoint:
+    description: Azure Trusted Signing endpoint.
+    required: true
+  account-name:
+    description: Azure Trusted Signing account name.
+    required: true
+  certificate-profile-name:
+    description: Certificate profile name for signing.
+    required: true
+
+runs:
+  using: composite
+  steps:
+    - name: Azure login for Trusted Signing (OIDC)
+      uses: azure/login@v2
+      with:
+        client-id: ${{ inputs.client-id }}
+        tenant-id: ${{ inputs.tenant-id }}
+        subscription-id: ${{ inputs.subscription-id }}
+
+    - name: Sign Windows binaries with Azure Trusted Signing
+      uses: azure/trusted-signing-action@v0
+      with:
+        endpoint: ${{ inputs.endpoint }}
+        trusted-signing-account-name: ${{ inputs.account-name }}
+        certificate-profile-name: ${{ inputs.certificate-profile-name }}
+        exclude-environment-credential: true
+        exclude-workload-identity-credential: true
+        exclude-managed-identity-credential: true
+        exclude-shared-token-cache-credential: true
+        exclude-visual-studio-credential: true
+        exclude-visual-studio-code-credential: true
+        exclude-azure-cli-credential: false
+        exclude-azure-powershell-credential: true
+        exclude-azure-developer-cli-credential: true
+        exclude-interactive-browser-credential: true
+        cache-dependencies: false
+        files: |
+          ${{ github.workspace }}/codex-rs/target/${{ inputs.target }}/release/codex.exe
+          ${{ github.workspace }}/codex-rs/target/${{ inputs.target }}/release/codex-responses-api-proxy.exe
--- a/.github/workflows/cla.yml
+++ b/.github/workflows/cla.yml
@@ -46,7 +46,4 @@ jobs:
          path-to-document: https://github.com/openai/codex/blob/main/docs/CLA.md
          path-to-signatures: signatures/cla.json
          branch: cla-signatures
-          allowlist: |
-            codex
-            dependabot
-            dependabot[bot]
+          allowlist: codex,dependabot,dependabot[bot],github-actions[bot]
--- a/.github/workflows/rust-ci.yml
+++ b/.github/workflows/rust-ci.yml
@@ -369,6 +369,49 @@ jobs:

    steps:
      - uses: actions/checkout@v6
+
+      # We have been running out of space when running this job on Linux for
+      # x86_64-unknown-linux-gnu, so remove some unnecessary dependencies.
+      - name: Remove unnecessary dependencies to save space
+        if: ${{ startsWith(matrix.runner, 'ubuntu') }}
+        shell: bash
+        run: |
+          set -euo pipefail
+          sudo rm -rf \
+            /usr/local/lib/android \
+            /usr/share/dotnet \
+            /usr/local/share/boost \
+            /usr/local/lib/node_modules \
+            /opt/ghc
+          sudo apt-get remove -y docker.io docker-compose podman buildah
+
+      # Ensure brew includes this fix so that brew's shellenv.sh loads
+      # cleanly in the Codex sandbox (it is frequently eval'd via .zprofile
+      # for Brew users, including the macOS runners on GitHub):
+      #
+      # https://github.com/Homebrew/brew/pull/21157
+      #
+      # Once brew 5.0.5 is released and is the default on macOS runners, this
+      # step can be removed.
+      - name: Upgrade brew
+        if: ${{ startsWith(matrix.runner, 'macos') }}
+        shell: bash
+        run: |
+          set -euo pipefail
+          brew --version
+          git -C "$(brew --repo)" fetch origin
+          git -C "$(brew --repo)" checkout main
+          git -C "$(brew --repo)" reset --hard origin/main
+          export HOMEBREW_UPDATE_TO_TAG=0
+          brew update
+          brew upgrade
+          brew --version
+
+      # Some integration tests rely on DotSlash being installed.
+      # See https://github.com/openai/codex/pull/7617.
+      - name: Install DotSlash
+        uses: facebook/install-dotslash@v2
+
      - uses: dtolnay/rust-toolchain@1.90
        with:
          targets: ${{ matrix.target }}
--- a/.github/workflows/rust-release.yml
+++ b/.github/workflows/rust-release.yml
@@ -50,6 +50,9 @@ jobs:
    name: Build - ${{ matrix.runner }} - ${{ matrix.target }}
    runs-on: ${{ matrix.runner }}
    timeout-minutes: 30
+    permissions:
+      contents: read
+      id-token: write
    defaults:
      run:
        working-directory: codex-rs
@@ -100,6 +103,25 @@ jobs:
      - name: Cargo build
        run: cargo build --target ${{ matrix.target }} --release --bin codex --bin codex-responses-api-proxy

+      - if: ${{ contains(matrix.target, 'linux') }}
+        name: Cosign Linux artifacts
+        uses: ./.github/actions/linux-code-sign
+        with:
+          target: ${{ matrix.target }}
+          artifacts-dir: ${{ github.workspace }}/codex-rs/target/${{ matrix.target }}/release
+
+      - if: ${{ contains(matrix.target, 'windows') }}
+        name: Sign Windows binaries with Azure Trusted Signing
+        uses: ./.github/actions/windows-code-sign
+        with:
+          target: ${{ matrix.target }}
+          client-id: ${{ secrets.AZURE_TRUSTED_SIGNING_CLIENT_ID }}
+          tenant-id: ${{ secrets.AZURE_TRUSTED_SIGNING_TENANT_ID }}
+          subscription-id: ${{ secrets.AZURE_TRUSTED_SIGNING_SUBSCRIPTION_ID }}
+          endpoint: ${{ secrets.AZURE_TRUSTED_SIGNING_ENDPOINT }}
+          account-name: ${{ secrets.AZURE_TRUSTED_SIGNING_ACCOUNT_NAME }}
+          certificate-profile-name: ${{ secrets.AZURE_TRUSTED_SIGNING_CERTIFICATE_PROFILE_NAME }}
+
      - if: ${{ matrix.runner == 'macos-15-xlarge' }}
        name: Configure Apple code signing
        shell: bash
@@ -283,6 +305,11 @@ jobs:
            cp target/${{ matrix.target }}/release/codex-responses-api-proxy "$dest/codex-responses-api-proxy-${{ matrix.target }}"
          fi

+          if [[ "${{ matrix.target }}" == *linux* ]]; then
+            cp target/${{ matrix.target }}/release/codex.sigstore "$dest/codex-${{ matrix.target }}.sigstore"
+            cp target/${{ matrix.target }}/release/codex-responses-api-proxy.sigstore "$dest/codex-responses-api-proxy-${{ matrix.target }}.sigstore"
+          fi
+
      - if: ${{ matrix.runner == 'windows-11-arm' }}
        name: Install zstd
        shell: powershell
@@ -321,6 +348,11 @@ jobs:
              continue
            fi

+            # Don't try to compress signature bundles.
+            if [[ "$base" == *.sigstore ]]; then
+              continue
+            fi
+
            # Create per-binary tar.gz
            tar -C "$dest" -czf "$dest/${base}.tar.gz" "$base"

--- a/AGENTS.md
+++ b/AGENTS.md
@@ -75,6 +75,7 @@ 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.
+- Prefer deep equals comparisons whenever possible. Perform `assert_eq!()` on entire objects, rather than individual fields.

 ### Integration tests (core)

--- a/codex-cli/bin/codex.js
+++ b/codex-cli/bin/codex.js
@@ -95,14 +95,6 @@ function detectPackageManager() {
    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;
 }

--- a/codex-rs/Cargo.lock
+++ b/codex-rs/Cargo.lock
@@ -198,9 +198,9 @@ dependencies = [

 [[package]]
 name = "arboard"
-version = "3.6.0"
+version = "3.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "55f533f8e0af236ffe5eb979b99381df3258853f00ba2e44b6e1955292c75227"
+checksum = "0348a1c054491f4bfe6ab86a7b6ab1e44e45d899005de92f58b3df180b36ddaf"
 dependencies = [
 "clipboard-win",
 "image",
@@ -212,7 +212,7 @@ dependencies = [
 "objc2-foundation",
 "parking_lot",
 "percent-encoding",
- "windows-sys 0.52.0",
+ "windows-sys 0.60.2",
 "wl-clipboard-rs",
 "x11rb",
 ]
@@ -238,48 +238,6 @@ dependencies = [
 "term",
 ]

-[[package]]
-name = "askama"
-version = "0.14.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f75363874b771be265f4ffe307ca705ef6f3baa19011c149da8674a87f1b75c4"
-dependencies = [
- "askama_derive",
- "itoa",
- "percent-encoding",
- "serde",
- "serde_json",
-]
-
-[[package]]
-name = "askama_derive"
-version = "0.14.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "129397200fe83088e8a68407a8e2b1f826cf0086b21ccdb866a722c8bcd3a94f"
-dependencies = [
- "askama_parser",
- "basic-toml",
- "memchr",
- "proc-macro2",
- "quote",
- "rustc-hash",
- "serde",
- "serde_derive",
- "syn 2.0.104",
-]
-
-[[package]]
-name = "askama_parser"
-version = "0.14.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d6ab5630b3d5eaf232620167977f95eb51f3432fc76852328774afbd242d4358"
-dependencies = [
- "memchr",
- "serde",
- "serde_derive",
- "winnow",
-]
-
 [[package]]
 name = "assert-json-diff"
 version = "2.0.2"
@@ -557,15 +515,6 @@ version = "0.22.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "72b3254f16251a8381aa12e40e3c4d2f0199f8c6508fbecb9d91f575e0fbb8c6"

-[[package]]
-name = "basic-toml"
-version = "0.1.10"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ba62675e8242a4c4e806d12f11d136e626e6c8361d6b829310732241652a178a"
-dependencies = [
- "serde",
-]
-
 [[package]]
 name = "beef"
 version = "0.5.2"
@@ -858,6 +807,7 @@ dependencies = [
 "http",
 "pretty_assertions",
 "regex-lite",
+ "reqwest",
 "serde",
 "serde_json",
 "thiserror 2.0.17",
@@ -865,6 +815,7 @@ dependencies = [
 "tokio-test",
 "tokio-util",
 "tracing",
+ "wiremock",
 ]

 [[package]]
@@ -885,6 +836,7 @@ dependencies = [
 "codex-file-search",
 "codex-login",
 "codex-protocol",
+ "codex-rmcp-client",
 "codex-utils-json-to-toml",
 "core_test_support",
 "mcp-types",
@@ -898,7 +850,8 @@ dependencies = [
 "shlex",
 "tempfile",
 "tokio",
- "toml",
+ "toml 0.9.5",
+ "toml_edit",
 "tracing",
 "tracing-subscriber",
 "uuid",
@@ -1038,6 +991,7 @@ dependencies = [
 "codex-rmcp-client",
 "codex-stdio-to-uds",
 "codex-tui",
+ "codex-tui2",
 "codex-windows-sandbox",
 "ctor 0.5.0",
 "libc",
@@ -1046,10 +1000,10 @@ dependencies = [
 "pretty_assertions",
 "regex-lite",
 "serde_json",
- "supports-color",
+ "supports-color 3.0.2",
 "tempfile",
 "tokio",
- "toml",
+ "toml 0.9.5",
 "tracing",
 ]

@@ -1068,6 +1022,7 @@ dependencies = [
 "serde_json",
 "thiserror 2.0.17",
 "tokio",
+ "tracing",
 ]

 [[package]]
@@ -1085,10 +1040,13 @@ dependencies = [
 "codex-login",
 "codex-tui",
 "crossterm",
+ "owo-colors",
+ "pretty_assertions",
 "ratatui",
 "reqwest",
 "serde",
 "serde_json",
+ "supports-color 3.0.2",
 "tokio",
 "tokio-stream",
 "tracing",
@@ -1116,14 +1074,12 @@ name = "codex-common"
 version = "0.0.0"
 dependencies = [
 "clap",
- "codex-app-server-protocol",
 "codex-core",
 "codex-lmstudio",
 "codex-ollama",
 "codex-protocol",
- "once_cell",
 "serde",
- "toml",
+ "toml 0.9.5",
 ]

 [[package]]
@@ -1131,7 +1087,6 @@ name = "codex-core"
 version = "0.0.0"
 dependencies = [
 "anyhow",
- "askama",
 "assert_cmd",
 "assert_matches",
 "async-channel",
@@ -1144,6 +1099,7 @@ dependencies = [
 "codex-apply-patch",
 "codex-arg0",
 "codex-async-utils",
+ "codex-client",
 "codex-core",
 "codex-execpolicy",
 "codex-file-search",
@@ -1152,6 +1108,7 @@ dependencies = [
 "codex-otel",
 "codex-protocol",
 "codex-rmcp-client",
+ "codex-utils-absolute-path",
 "codex-utils-pty",
 "codex-utils-readiness",
 "codex-utils-string",
@@ -1186,6 +1143,7 @@ dependencies = [
 "seccompiler",
 "serde",
 "serde_json",
+ "serde_yaml",
 "serial_test",
 "sha1",
 "sha2",
@@ -1200,7 +1158,7 @@ dependencies = [
 "tokio",
 "tokio-test",
 "tokio-util",
- "toml",
+ "toml 0.9.5",
 "toml_edit",
 "tracing",
 "tracing-test",
@@ -1235,7 +1193,7 @@ dependencies = [
 "serde",
 "serde_json",
 "shlex",
- "supports-color",
+ "supports-color 3.0.2",
 "tempfile",
 "tokio",
 "tracing",
@@ -1251,10 +1209,14 @@ name = "codex-exec-server"
 version = "0.0.0"
 dependencies = [
 "anyhow",
+ "assert_cmd",
 "async-trait",
 "clap",
 "codex-core",
+ "codex-execpolicy",
+ "exec_server_test_support",
 "libc",
+ "maplit",
 "path-absolutize",
 "pretty_assertions",
 "rmcp",
@@ -1267,6 +1229,7 @@ dependencies = [
 "tokio-util",
 "tracing",
 "tracing-subscriber",
+ "which",
 ]

 [[package]]
@@ -1281,6 +1244,7 @@ dependencies = [
 "serde_json",
 "shlex",
 "starlark",
+ "tempfile",
 "thiserror 2.0.17",
 ]

@@ -1291,7 +1255,7 @@ dependencies = [
 "allocative",
 "anyhow",
 "clap",
- "derive_more 2.0.1",
+ "derive_more 2.1.0",
 "env_logger",
 "log",
 "multimap",
@@ -1452,6 +1416,7 @@ dependencies = [
 "chrono",
 "codex-app-server-protocol",
 "codex-protocol",
+ "codex-utils-absolute-path",
 "eventsource-stream",
 "http",
 "opentelemetry",
@@ -1472,6 +1437,7 @@ name = "codex-process-hardening"
 version = "0.0.0"
 dependencies = [
 "libc",
+ "pretty_assertions",
 ]

 [[package]]
@@ -1479,7 +1445,6 @@ name = "codex-protocol"
 version = "0.0.0"
 dependencies = [
 "anyhow",
- "base64",
 "codex-git",
 "codex-utils-image",
 "icu_decimal",
@@ -1582,7 +1547,7 @@ dependencies = [
 "codex-windows-sandbox",
 "color-eyre",
 "crossterm",
- "derive_more 2.0.1",
+ "derive_more 2.1.0",
 "diffy",
 "dirs",
 "dunce",
@@ -1607,12 +1572,13 @@ dependencies = [
 "shlex",
 "strum 0.27.2",
 "strum_macros 0.27.2",
- "supports-color",
+ "supports-color 3.0.2",
 "tempfile",
 "textwrap 0.16.2",
 "tokio",
 "tokio-stream",
- "toml",
+ "tokio-util",
+ "toml 0.9.5",
 "tracing",
 "tracing-appender",
 "tracing-subscriber",
@@ -1621,9 +1587,89 @@ dependencies = [
 "unicode-segmentation",
 "unicode-width 0.2.1",
 "url",
+ "uuid",
 "vt100",
 ]

+[[package]]
+name = "codex-tui2"
+version = "0.0.0"
+dependencies = [
+ "anyhow",
+ "arboard",
+ "assert_matches",
+ "async-stream",
+ "base64",
+ "chrono",
+ "clap",
+ "codex-ansi-escape",
+ "codex-app-server-protocol",
+ "codex-arg0",
+ "codex-backend-client",
+ "codex-common",
+ "codex-core",
+ "codex-feedback",
+ "codex-file-search",
+ "codex-login",
+ "codex-protocol",
+ "codex-tui",
+ "codex-windows-sandbox",
+ "color-eyre",
+ "crossterm",
+ "derive_more 2.1.0",
+ "diffy",
+ "dirs",
+ "dunce",
+ "image",
+ "insta",
+ "itertools 0.14.0",
+ "lazy_static",
+ "libc",
+ "mcp-types",
+ "opentelemetry-appender-tracing",
+ "pathdiff",
+ "pretty_assertions",
+ "pulldown-cmark",
+ "rand 0.9.2",
+ "ratatui",
+ "ratatui-macros",
+ "regex-lite",
+ "reqwest",
+ "serde",
+ "serde_json",
+ "serial_test",
+ "shlex",
+ "strum 0.27.2",
+ "strum_macros 0.27.2",
+ "supports-color 3.0.2",
+ "tempfile",
+ "textwrap 0.16.2",
+ "tokio",
+ "tokio-stream",
+ "tokio-util",
+ "toml 0.9.5",
+ "tracing",
+ "tracing-appender",
+ "tracing-subscriber",
+ "tree-sitter-bash",
+ "tree-sitter-highlight",
+ "unicode-segmentation",
+ "unicode-width 0.2.1",
+ "url",
+ "uuid",
+ "vt100",
+]
+
+[[package]]
+name = "codex-utils-absolute-path"
+version = "0.0.0"
+dependencies = [
+ "path-absolutize",
+ "serde",
+ "serde_json",
+ "tempfile",
+]
+
 [[package]]
 name = "codex-utils-cache"
 version = "0.0.0"
@@ -1651,7 +1697,7 @@ version = "0.0.0"
 dependencies = [
 "pretty_assertions",
 "serde_json",
- "toml",
+ "toml 0.9.5",
 ]

 [[package]]
@@ -1659,8 +1705,13 @@ name = "codex-utils-pty"
 version = "0.0.0"
 dependencies = [
 "anyhow",
+ "filedescriptor",
+ "lazy_static",
+ "log",
 "portable-pty",
+ "shared_library",
 "tokio",
+ "winapi",
 ]

 [[package]]
@@ -1683,6 +1734,8 @@ name = "codex-windows-sandbox"
 version = "0.0.0"
 dependencies = [
 "anyhow",
+ "base64",
+ "chrono",
 "codex-protocol",
 "dirs-next",
 "dunce",
@@ -1690,7 +1743,9 @@ dependencies = [
 "serde",
 "serde_json",
 "tempfile",
+ "windows 0.58.0",
 "windows-sys 0.52.0",
+ "winres",
 ]

 [[package]]
@@ -1782,9 +1837,9 @@ dependencies = [

 [[package]]
 name = "convert_case"
-version = "0.7.1"
+version = "0.10.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bb402b8d4c85569410425650ce3eddc7d698ed96d39a73f941b08fb63082f1e7"
+checksum = "633458d4ef8c78b72454de2d54fd6ab2e60f9e02be22f3c6104cdc8a4e0fceb9"
 dependencies = [
 "unicode-segmentation",
 ]
@@ -2123,11 +2178,11 @@ dependencies = [

 [[package]]
 name = "derive_more"
-version = "2.0.1"
+version = "2.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "093242cf7570c207c83073cf82f79706fe7b8317e98620a47d5be7c3d8497678"
+checksum = "10b768e943bed7bf2cab53df09f4bc34bfd217cdb57d971e769874c9a6710618"
 dependencies = [
- "derive_more-impl 2.0.1",
+ "derive_more-impl 2.1.0",
 ]

 [[package]]
@@ -2145,13 +2200,14 @@ dependencies = [

 [[package]]
 name = "derive_more-impl"
-version = "2.0.1"
+version = "2.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bda628edc44c4bb645fbe0f758797143e4e07926f7ebf4e9bdfbd3d2ce621df3"
+checksum = "6d286bfdaf75e988b4a78e013ecd79c581e06399ab53fbacd2d916c2f904f30b"
 dependencies = [
- "convert_case 0.7.1",
+ "convert_case 0.10.0",
 "proc-macro2",
 "quote",
+ "rustc_version",
 "syn 2.0.104",
 "unicode-xid",
 ]
@@ -2492,6 +2548,18 @@ dependencies = [
 "pin-project-lite",
 ]

+[[package]]
+name = "exec_server_test_support"
+version = "0.0.0"
+dependencies = [
+ "anyhow",
+ "assert_cmd",
+ "codex-core",
+ "rmcp",
+ "serde_json",
+ "tokio",
+]
+
 [[package]]
 name = "eyre"
 version = "0.6.12"
@@ -2536,7 +2604,7 @@ checksum = "0ce92ff622d6dadf7349484f42c93271a0d49b7cc4d466a936405bacbe10aa78"
 dependencies = [
 "cfg-if",
 "rustix 1.0.8",
- "windows-sys 0.52.0",
+ "windows-sys 0.59.0",
 ]

 [[package]]
@@ -3110,7 +3178,7 @@ dependencies = [
 "js-sys",
 "log",
 "wasm-bindgen",
- "windows-core",
+ "windows-core 0.61.2",
 ]

 [[package]]
@@ -3293,9 +3361,9 @@ dependencies = [

 [[package]]
 name = "image"
-version = "0.25.8"
+version = "0.25.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "529feb3e6769d234375c4cf1ee2ce713682b8e76538cb13f9fc23e1400a591e7"
+checksum = "e6506c6c10786659413faa717ceebcb8f70731c0a60cbae39795fdf114519c1a"
 dependencies = [
 "bytemuck",
 "byteorder-lite",
@@ -3303,8 +3371,8 @@ dependencies = [
 "num-traits",
 "png",
 "tiff",
- "zune-core",
- "zune-jpeg",
+ "zune-core 0.5.0",
+ "zune-jpeg 0.5.5",
 ]

 [[package]]
@@ -3374,9 +3442,9 @@ dependencies = [

 [[package]]
 name = "insta"
-version = "1.43.2"
+version = "1.44.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "46fdb647ebde000f43b5b53f773c30cf9b0cb4300453208713fa38b2c70935a0"
+checksum = "b5c943d4415edd8153251b6f197de5eb1640e56d84e8d9159bea190421c73698"
 dependencies = [
 "console",
 "once_cell",
@@ -3440,7 +3508,7 @@ checksum = "e04d7f318608d35d4b61ddd75cbdaee86b023ebe2bd5a66ee0915f0bf93095a9"
 dependencies = [
 "hermit-abi",
 "libc",
- "windows-sys 0.52.0",
+ "windows-sys 0.59.0",
 ]

 [[package]]
@@ -4428,6 +4496,10 @@ name = "owo-colors"
 version = "4.2.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "48dd4f4a2c8405440fd0462561f0e5806bd0f77e86f51c761481bdd4018b545e"
+dependencies = [
+ "supports-color 2.1.0",
+ "supports-color 3.0.2",
+]

 [[package]]
 name = "parking"
@@ -4464,6 +4536,12 @@ version = "1.0.15"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "57c0d7b74b563b49d38dae00a0c37d4d6de9b432382b2892f0574ddcae73fd0a"

+[[package]]
+name = "pastey"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "57d6c094ee800037dff99e02cab0eaf3142826586742a270ab3d7a62656bd27a"
+
 [[package]]
 name = "path-absolutize"
 version = "3.1.1"
@@ -4739,7 +4817,7 @@ dependencies = [
 "nix 0.30.1",
 "tokio",
 "tracing",
- "windows",
+ "windows 0.61.3",
 ]

 [[package]]
@@ -5079,9 +5157,9 @@ checksum = "2b15c43186be67a4fd63bee50d0303afffcef381492ebe2c5d87f324e1b8815c"

 [[package]]
 name = "reqwest"
-version = "0.12.23"
+version = "0.12.24"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d429f34c8092b2d42c7c93cec323bb4adeb7c67698f70839adec842ec10c7ceb"
+checksum = "9d0946410b9f7b082a427e4ef5c8ff541a88b357bc6c637c40db3a68ac70a36f"
 dependencies = [
 "base64",
 "bytes",
@@ -5142,8 +5220,9 @@ dependencies = [

 [[package]]
 name = "rmcp"
-version = "0.9.0"
-source = "git+https://github.com/bolinfest/rust-sdk?branch=pr556#4d9cc16f4c76c84486344f542ed9a3e9364019ba"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "38b18323edc657390a6ed4d7a9110b0dec2dc3ed128eb2a123edfbafabdbddc5"
 dependencies = [
 "async-trait",
 "base64",
@@ -5154,7 +5233,7 @@ dependencies = [
 "http-body",
 "http-body-util",
 "oauth2",
- "paste",
+ "pastey",
 "pin-project-lite",
 "process-wrap",
 "rand 0.9.2",
@@ -5176,8 +5255,9 @@ dependencies = [

 [[package]]
 name = "rmcp-macros"
-version = "0.9.0"
-source = "git+https://github.com/bolinfest/rust-sdk?branch=pr556#4d9cc16f4c76c84486344f542ed9a3e9364019ba"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "c75d0a62676bf8c8003c4e3c348e2ceb6a7b3e48323681aaf177fdccdac2ce50"
 dependencies = [
 "darling 0.21.3",
 "proc-macro2",
@@ -5217,7 +5297,7 @@ dependencies = [
 "errno",
 "libc",
 "linux-raw-sys 0.4.15",
- "windows-sys 0.52.0",
+ "windows-sys 0.59.0",
 ]

 [[package]]
@@ -5736,9 +5816,9 @@ dependencies = [

 [[package]]
 name = "serde_with"
-version = "3.14.0"
+version = "3.16.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f2c45cd61fefa9db6f254525d46e392b852e0e61d9a1fd36e5bd183450a556d5"
+checksum = "4fa237f2807440d238e0364a218270b98f767a00d3dada77b1c53ae88940e2e7"
 dependencies = [
 "base64",
 "chrono",
@@ -5747,8 +5827,7 @@ dependencies = [
 "indexmap 2.12.0",
 "schemars 0.9.0",
 "schemars 1.0.4",
- "serde",
- "serde_derive",
+ "serde_core",
 "serde_json",
 "serde_with_macros",
 "time",
@@ -5756,16 +5835,29 @@ dependencies = [

 [[package]]
 name = "serde_with_macros"
-version = "3.14.0"
+version = "3.16.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "de90945e6565ce0d9a25098082ed4ee4002e047cb59892c318d66821e14bb30f"
+checksum = "52a8e3ca0ca629121f70ab50f95249e5a6f925cc0f6ffe8256c45b728875706c"
 dependencies = [
- "darling 0.20.11",
+ "darling 0.21.3",
 "proc-macro2",
 "quote",
 "syn 2.0.104",
 ]

+[[package]]
+name = "serde_yaml"
+version = "0.9.34+deprecated"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6a8b1a1a2ebf674015cc02edccce75287f1a0130d394307b36743c2f5d504b47"
+dependencies = [
+ "indexmap 2.12.0",
+ "itoa",
+ "ryu",
+ "serde",
+ "unsafe-libyaml",
+]
+
 [[package]]
 name = "serial2"
 version = "0.2.31"
@@ -6144,6 +6236,16 @@ version = "2.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "13c2bddecc57b384dee18652358fb23172facb8a2c51ccc10d74c157bdea3292"

+[[package]]
+name = "supports-color"
+version = "2.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "d6398cde53adc3c4557306a96ce67b302968513830a77a95b2b17305d9719a89"
+dependencies = [
+ "is-terminal",
+ "is_ci",
+]
+
 [[package]]
 name = "supports-color"
 version = "3.0.2"
@@ -6409,7 +6511,7 @@ dependencies = [
 "half",
 "quick-error",
 "weezl",
- "zune-jpeg",
+ "zune-jpeg 0.4.19",
 ]

 [[package]]
@@ -6577,9 +6679,19 @@ dependencies = [
 "futures-sink",
 "futures-util",
 "pin-project-lite",
+ "slab",
 "tokio",
 ]

+[[package]]
+name = "toml"
+version = "0.5.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f4f7f0dd8d50a853a531c426359045b1998f04219d88799810762cd4ad314234"
+dependencies = [
+ "serde",
+]
+
 [[package]]
 name = "toml"
 version = "0.9.5"
@@ -6714,9 +6826,9 @@ checksum = "8df9b6e13f2d32c91b9bd719c00d1958837bc7dec474d94952798cc8e69eeec3"

 [[package]]
 name = "tracing"
-version = "0.1.41"
+version = "0.1.43"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "784e0ac535deb450455cbfa28a6f0df145ea1bb7ae51b821cf5e7927fdcfbdd0"
+checksum = "2d15d90a0b5c19378952d479dc858407149d7bb45a14de0142f6c534b16fc647"
 dependencies = [
 "log",
 "pin-project-lite",
@@ -6738,9 +6850,9 @@ dependencies = [

 [[package]]
 name = "tracing-attributes"
-version = "0.1.30"
+version = "0.1.31"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "81383ab64e72a7a8b8e13130c49e3dab29def6d0c7d76a03087b3cf71c5c6903"
+checksum = "7490cfa5ec963746568740651ac6781f701c9c5ea257c58e057f3ba8cf69e8da"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -6749,9 +6861,9 @@ dependencies = [

 [[package]]
 name = "tracing-core"
-version = "0.1.34"
+version = "0.1.35"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b9d12581f227e93f094d3af2ae690a574abb8a2b9b7a96e7cfe9647b2b617678"
+checksum = "7a04e24fab5c89c6a36eb8558c9656f30d81de51dfa4d3b45f26b21d61fa0a6c"
 dependencies = [
 "once_cell",
 "valuable",
@@ -6879,9 +6991,9 @@ checksum = "e421abadd41a4225275504ea4d6566923418b7f05506fbc9c0fe86ba7396114b"

 [[package]]
 name = "ts-rs"
-version = "11.0.1"
+version = "11.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6ef1b7a6d914a34127ed8e1fa927eb7088903787bcded4fa3eef8f85ee1568be"
+checksum = "4994acea2522cd2b3b85c1d9529a55991e3ad5e25cdcd3de9d505972c4379424"
 dependencies = [
 "serde_json",
 "thiserror 2.0.17",
@@ -6891,9 +7003,9 @@ dependencies = [

 [[package]]
 name = "ts-rs-macros"
-version = "11.0.1"
+version = "11.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "e9d4ed7b4c18cc150a6a0a1e9ea1ecfa688791220781af6e119f9599a8502a0a"
+checksum = "ee6ff59666c9cbaec3533964505d39154dc4e0a56151fdea30a09ed0301f62e2"
 dependencies = [
 "proc-macro2",
 "quote",
@@ -6980,6 +7092,12 @@ version = "0.2.6"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "ebc1c04c71510c7f702b52b7c350734c9ff1295c464a03335b00bb84fc54f853"

+[[package]]
+name = "unsafe-libyaml"
+version = "0.2.11"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "673aac59facbab8a9007c7f6108d11f63b603f7cabff99fabf650fea5c32b861"
+
 [[package]]
 name = "untrusted"
 version = "0.9.0"
@@ -7343,9 +7461,9 @@ dependencies = [

 [[package]]
 name = "wildmatch"
-version = "2.5.0"
+version = "2.6.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "39b7d07a236abaef6607536ccfaf19b396dbe3f5110ddb73d39f4562902ed382"
+checksum = "29333c3ea1ba8b17211763463ff24ee84e41c78224c16b001cd907e663a38c68"

 [[package]]
 name = "winapi"
@@ -7369,7 +7487,7 @@ version = "0.1.9"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "cf221c93e13a30d793f7645a0e7762c55d169dbb0a49671918a2319d289b10bb"
 dependencies = [
- "windows-sys 0.52.0",
+ "windows-sys 0.59.0",
 ]

 [[package]]
@@ -7378,6 +7496,16 @@ version = "0.4.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"

+[[package]]
+name = "windows"
+version = "0.58.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dd04d41d93c4992d421894c18c8b43496aa748dd4c081bac0dc93eb0489272b6"
+dependencies = [
+ "windows-core 0.58.0",
+ "windows-targets 0.52.6",
+]
+
 [[package]]
 name = "windows"
 version = "0.61.3"
@@ -7385,7 +7513,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9babd3a767a4c1aef6900409f85f5d53ce2544ccdfaa86dad48c91782c6d6893"
 dependencies = [
 "windows-collections",
- "windows-core",
+ "windows-core 0.61.2",
 "windows-future",
 "windows-link 0.1.3",
 "windows-numerics",
@@ -7397,7 +7525,20 @@ version = "0.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3beeceb5e5cfd9eb1d76b381630e82c4241ccd0d27f1a39ed41b2760b255c5e8"
 dependencies = [
- "windows-core",
+ "windows-core 0.61.2",
+]
+
+[[package]]
+name = "windows-core"
+version = "0.58.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "6ba6d44ec8c2591c134257ce647b7ea6b20335bf6379a27dac5f1641fcf59f99"
+dependencies = [
+ "windows-implement 0.58.0",
+ "windows-interface 0.58.0",
+ "windows-result 0.2.0",
+ "windows-strings 0.1.0",
+ "windows-targets 0.52.6",
 ]

 [[package]]
@@ -7406,11 +7547,11 @@ version = "0.61.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "c0fdd3ddb90610c7638aa2b3a3ab2904fb9e5cdbecc643ddb3647212781c4ae3"
 dependencies = [
- "windows-implement",
- "windows-interface",
+ "windows-implement 0.60.0",
+ "windows-interface 0.59.1",
 "windows-link 0.1.3",
- "windows-result",
- "windows-strings",
+ "windows-result 0.3.4",
+ "windows-strings 0.4.2",
 ]

 [[package]]
@@ -7419,11 +7560,22 @@ version = "0.2.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "fc6a41e98427b19fe4b73c550f060b59fa592d7d686537eebf9385621bfbad8e"
 dependencies = [
- "windows-core",
+ "windows-core 0.61.2",
 "windows-link 0.1.3",
 "windows-threading",
 ]

+[[package]]
+name = "windows-implement"
+version = "0.58.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2bbd5b46c938e506ecbce286b6628a02171d56153ba733b6c741fc627ec9579b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.104",
+]
+
 [[package]]
 name = "windows-implement"
 version = "0.60.0"
@@ -7435,6 +7587,17 @@ dependencies = [
 "syn 2.0.104",
 ]

+[[package]]
+name = "windows-interface"
+version = "0.58.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "053c4c462dc91d3b1504c6fe5a726dd15e216ba718e84a0e46a88fbe5ded3515"
+dependencies = [
+ "proc-macro2",
+ "quote",
+ "syn 2.0.104",
+]
+
 [[package]]
 name = "windows-interface"
 version = "0.59.1"
@@ -7464,7 +7627,7 @@ version = "0.2.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9150af68066c4c5c07ddc0ce30421554771e528bde427614c61038bc2c92c2b1"
 dependencies = [
- "windows-core",
+ "windows-core 0.61.2",
 "windows-link 0.1.3",
 ]

@@ -7475,8 +7638,17 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "5b8a9ed28765efc97bbc954883f4e6796c33a06546ebafacbabee9696967499e"
 dependencies = [
 "windows-link 0.1.3",
- "windows-result",
- "windows-strings",
+ "windows-result 0.3.4",
+ "windows-strings 0.4.2",
+]
+
+[[package]]
+name = "windows-result"
+version = "0.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "1d1043d8214f791817bab27572aaa8af63732e11bf84aa21a45a78d6c317ae0e"
+dependencies = [
+ "windows-targets 0.52.6",
 ]

 [[package]]
@@ -7488,6 +7660,16 @@ dependencies = [
 "windows-link 0.1.3",
 ]

+[[package]]
+name = "windows-strings"
+version = "0.1.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "4cd9b125c486025df0eabcb585e62173c6c9eddcec5d117d3b6e8c30e2ee4d10"
+dependencies = [
+ "windows-result 0.2.0",
+ "windows-targets 0.52.6",
+]
+
 [[package]]
 name = "windows-strings"
 version = "0.4.2"
@@ -7811,6 +7993,15 @@ dependencies = [
 "winapi",
 ]

+[[package]]
+name = "winres"
+version = "0.1.12"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "b68db261ef59e9e52806f688020631e987592bd83619edccda9c47d42cde4f6c"
+dependencies = [
+ "toml 0.5.11",
+]
+
 [[package]]
 name = "winsafe"
 version = "0.0.19"
@@ -8094,13 +8285,28 @@ version = "0.4.12"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3f423a2c17029964870cfaabb1f13dfab7d092a62a29a89264f4d36990ca414a"

+[[package]]
+name = "zune-core"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "111f7d9820f05fd715df3144e254d6fc02ee4088b0644c0ffd0efc9e6d9d2773"
+
 [[package]]
 name = "zune-jpeg"
 version = "0.4.19"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "2c9e525af0a6a658e031e95f14b7f889976b74a11ba0eca5a5fc9ac8a1c43a6a"
 dependencies = [
- "zune-core",
+ "zune-core 0.4.12",
+]
+
+[[package]]
+name = "zune-jpeg"
+version = "0.5.5"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "dc6fb7703e32e9a07fb3f757360338b3a567a5054f21b5f52a666752e333d58e"
+dependencies = [
+ "zune-core 0.5.0",
 ]

 [[package]]
--- a/codex-rs/Cargo.toml
+++ b/codex-rs/Cargo.toml
@@ -34,6 +34,8 @@ members = [
    "stdio-to-uds",
    "otel",
    "tui",
+    "tui2",
+    "utils/absolute-path",
    "utils/git",
    "utils/cache",
    "utils/image",
@@ -59,15 +61,15 @@ license = "Apache-2.0"
 # Internal
 app_test_support = { path = "app-server/tests/common" }
 codex-ansi-escape = { path = "ansi-escape" }
+codex-api = { path = "codex-api" }
 codex-app-server = { path = "app-server" }
 codex-app-server-protocol = { path = "app-server-protocol" }
 codex-apply-patch = { path = "apply-patch" }
 codex-arg0 = { path = "arg0" }
 codex-async-utils = { path = "async-utils" }
 codex-backend-client = { path = "backend-client" }
-codex-api = { path = "codex-api" }
-codex-client = { path = "codex-client" }
 codex-chatgpt = { path = "chatgpt" }
+codex-client = { path = "codex-client" }
 codex-common = { path = "common" }
 codex-core = { path = "core" }
 codex-exec = { path = "exec" }
@@ -88,6 +90,8 @@ codex-responses-api-proxy = { path = "responses-api-proxy" }
 codex-rmcp-client = { path = "rmcp-client" }
 codex-stdio-to-uds = { path = "stdio-to-uds" }
 codex-tui = { path = "tui" }
+codex-tui2 = { path = "tui2" }
+codex-utils-absolute-path = { path = "utils/absolute-path" }
 codex-utils-cache = { path = "utils/cache" }
 codex-utils-image = { path = "utils/image" }
 codex-utils-json-to-toml = { path = "utils/json-to-toml" }
@@ -96,6 +100,7 @@ codex-utils-readiness = { path = "utils/readiness" }
 codex-utils-string = { path = "utils/string" }
 codex-windows-sandbox = { path = "windows-sandbox-rs" }
 core_test_support = { path = "core/tests/common" }
+exec_server_test_support = { path = "exec-server/tests/common" }
 mcp-types = { path = "mcp-types" }
 mcp_test_support = { path = "mcp-server/tests/common" }

@@ -104,7 +109,6 @@ allocative = "0.3.3"
 ansi-to-tui = "7.0.0"
 anyhow = "1"
 arboard = { version = "3", features = ["wayland-data-control"] }
-askama = "0.14"
 assert_cmd = "2"
 assert_matches = "1.5.0"
 async-channel = "2.3.1"
@@ -136,9 +140,9 @@ icu_decimal = "2.1"
 icu_locale_core = "2.1"
 icu_provider = { version = "2.1", features = ["sync"] }
 ignore = "0.4.23"
-image = { version = "^0.25.8", default-features = false }
+image = { version = "^0.25.9", default-features = false }
 indexmap = "2.12.0"
-insta = "1.43.2"
+insta = "1.44.3"
 itertools = "0.14.0"
 keyring = { version = "3.6", default-features = false }
 landlock = "0.4.1"
@@ -169,16 +173,17 @@ pulldown-cmark = "0.10"
 rand = "0.9"
 ratatui = "0.29.0"
 ratatui-macros = "0.6.0"
-regex-lite = "0.1.7"
 regex = "1.12.2"
+regex-lite = "0.1.7"
 reqwest = "0.12"
-rmcp = { version = "0.9.0", default-features = false }
+rmcp = { version = "0.10.0", default-features = false }
 schemars = "0.8.22"
 seccompiler = "0.5.0"
 sentry = "0.34.0"
 serde = "1"
 serde_json = "1"
-serde_with = "3.14"
+serde_with = "3.16"
+serde_yaml = "0.9"
 serial_test = "3.2.0"
 sha1 = "0.10.6"
 sha2 = "0.10"
@@ -203,7 +208,7 @@ tokio-util = "0.7.16"
 toml = "0.9.5"
 toml_edit = "0.23.5"
 tonic = "0.13.1"
-tracing = "0.1.41"
+tracing = "0.1.43"
 tracing-appender = "0.2.3"
 tracing-subscriber = "0.3.20"
 tracing-test = "0.2.5"
@@ -221,7 +226,7 @@ vt100 = "0.16.2"
 walkdir = "2.5.0"
 webbrowser = "1.0"
 which = "6"
-wildmatch = "2.5.0"
+wildmatch = "2.6.1"

 wiremock = "0.6"
 zeroize = "1.8.2"
@@ -288,7 +293,6 @@ opt-level = 0
 # ratatui = { path = "../../ratatui" }
 crossterm = { git = "https://github.com/nornagon/crossterm", branch = "nornagon/color-query" }
 ratatui = { git = "https://github.com/nornagon/ratatui", branch = "nornagon-v0.29.0-patch" }
-rmcp = { git = "https://github.com/bolinfest/rust-sdk", branch = "pr556" }

 # Uncomment to debug local changes.
 # rmcp = { path = "../../rust-sdk/crates/rmcp" }
--- a/codex-rs/app-server-protocol/src/protocol/common.rs
+++ b/codex-rs/app-server-protocol/src/protocol/common.rs
@@ -131,7 +131,7 @@ client_request_definitions! {
    },
    ReviewStart => "review/start" {
        params: v2::ReviewStartParams,
-        response: v2::TurnStartResponse,
+        response: v2::ReviewStartResponse,
    },

    ModelList => "model/list" {
@@ -139,6 +139,16 @@ client_request_definitions! {
        response: v2::ModelListResponse,
    },

+    McpServerOauthLogin => "mcpServer/oauth/login" {
+        params: v2::McpServerOauthLoginParams,
+        response: v2::McpServerOauthLoginResponse,
+    },
+
+    McpServersList => "mcpServers/list" {
+        params: v2::ListMcpServersParams,
+        response: v2::ListMcpServersResponse,
+    },
+
    LoginAccount => "account/login/start" {
        params: v2::LoginAccountParams,
        response: v2::LoginAccountResponse,
@@ -164,6 +174,12 @@ client_request_definitions! {
        response: v2::FeedbackUploadResponse,
    },

+    /// Execute a command (argv vector) under the server's sandbox.
+    OneOffCommandExec => "command/exec" {
+        params: v2::CommandExecParams,
+        response: v2::CommandExecResponse,
+    },
+
    ConfigRead => "config/read" {
        params: v2::ConfigReadParams,
        response: v2::ConfigReadResponse,
@@ -506,11 +522,15 @@ server_notification_definitions! {
    TurnStarted => "turn/started" (v2::TurnStartedNotification),
    TurnCompleted => "turn/completed" (v2::TurnCompletedNotification),
    TurnDiffUpdated => "turn/diff/updated" (v2::TurnDiffUpdatedNotification),
+    TurnPlanUpdated => "turn/plan/updated" (v2::TurnPlanUpdatedNotification),
    ItemStarted => "item/started" (v2::ItemStartedNotification),
    ItemCompleted => "item/completed" (v2::ItemCompletedNotification),
    AgentMessageDelta => "item/agentMessage/delta" (v2::AgentMessageDeltaNotification),
    CommandExecutionOutputDelta => "item/commandExecution/outputDelta" (v2::CommandExecutionOutputDeltaNotification),
+    TerminalInteraction => "item/commandExecution/terminalInteraction" (v2::TerminalInteractionNotification),
+    FileChangeOutputDelta => "item/fileChange/outputDelta" (v2::FileChangeOutputDeltaNotification),
    McpToolCallProgress => "item/mcpToolCall/progress" (v2::McpToolCallProgressNotification),
+    McpServerOauthLoginCompleted => "mcpServer/oauthLogin/completed" (v2::McpServerOauthLoginCompletedNotification),
    AccountUpdated => "account/updated" (v2::AccountUpdatedNotification),
    AccountRateLimitsUpdated => "account/rateLimits/updated" (v2::AccountRateLimitsUpdatedNotification),
    ReasoningSummaryTextDelta => "item/reasoning/summaryTextDelta" (v2::ReasoningSummaryTextDeltaNotification),
@@ -634,7 +654,6 @@ mod tests {
            command: vec!["echo".to_string(), "hello".to_string()],
            cwd: PathBuf::from("/tmp"),
            reason: Some("because tests".to_string()),
-            risk: None,
            parsed_cmd: vec![ParsedCommand::Unknown {
                cmd: "echo hello".to_string(),
            }],
@@ -654,7 +673,6 @@ mod tests {
                    "command": ["echo", "hello"],
                    "cwd": "/tmp",
                    "reason": "because tests",
-                    "risk": null,
                    "parsedCmd": [
                        {
                            "type": "unknown",
--- a/codex-rs/app-server-protocol/src/protocol/mappers.rs
+++ b/codex-rs/app-server-protocol/src/protocol/mappers.rs
@@ -0,0 +1,15 @@
+use crate::protocol::v1;
+use crate::protocol::v2;
+
+impl From<v1::ExecOneOffCommandParams> for v2::CommandExecParams {
+    fn from(value: v1::ExecOneOffCommandParams) -> Self {
+        Self {
+            command: value.command,
+            timeout_ms: value
+                .timeout_ms
+                .map(|timeout| i64::try_from(timeout).unwrap_or(60_000)),
+            cwd: value.cwd,
+            sandbox_policy: value.sandbox_policy.map(std::convert::Into::into),
+        }
+    }
+}
--- a/codex-rs/app-server-protocol/src/protocol/mod.rs
+++ b/codex-rs/app-server-protocol/src/protocol/mod.rs
@@ -2,6 +2,7 @@
 // Exposes protocol pieces used by `lib.rs` via `pub use protocol::common::*;`.

 pub mod common;
+mod mappers;
 pub mod thread_history;
 pub mod v1;
 pub mod v2;
--- a/codex-rs/app-server-protocol/src/protocol/thread_history.rs
+++ b/codex-rs/app-server-protocol/src/protocol/thread_history.rs
@@ -1,5 +1,6 @@
 use crate::protocol::v2::ThreadItem;
 use crate::protocol::v2::Turn;
+use crate::protocol::v2::TurnError;
 use crate::protocol::v2::TurnStatus;
 use crate::protocol::v2::UserInput;
 use codex_protocol::protocol::AgentReasoningEvent;
@@ -142,6 +143,7 @@ impl ThreadHistoryBuilder {
        PendingTurn {
            id: self.next_turn_id(),
            items: Vec::new(),
+            error: None,
            status: TurnStatus::Completed,
        }
    }
@@ -190,6 +192,7 @@ impl ThreadHistoryBuilder {
 struct PendingTurn {
    id: String,
    items: Vec<ThreadItem>,
+    error: Option<TurnError>,
    status: TurnStatus,
 }

@@ -198,6 +201,7 @@ impl From<PendingTurn> for Turn {
        Self {
            id: value.id,
            items: value.items,
+            error: value.error,
            status: value.status,
        }
    }
--- a/codex-rs/app-server-protocol/src/protocol/v1.rs
+++ b/codex-rs/app-server-protocol/src/protocol/v1.rs
@@ -3,17 +3,16 @@ use std::path::PathBuf;

 use codex_protocol::ConversationId;
 use codex_protocol::config_types::ForcedLoginMethod;
-use codex_protocol::config_types::ReasoningEffort;
 use codex_protocol::config_types::ReasoningSummary;
 use codex_protocol::config_types::SandboxMode;
 use codex_protocol::config_types::Verbosity;
 use codex_protocol::models::ResponseItem;
+use codex_protocol::openai_models::ReasoningEffort;
 use codex_protocol::parse_command::ParsedCommand;
 use codex_protocol::protocol::AskForApproval;
 use codex_protocol::protocol::EventMsg;
 use codex_protocol::protocol::FileChange;
 use codex_protocol::protocol::ReviewDecision;
-use codex_protocol::protocol::SandboxCommandAssessment;
 use codex_protocol::protocol::SandboxPolicy;
 use codex_protocol::protocol::SessionSource;
 use codex_protocol::protocol::TurnAbortReason;
@@ -226,7 +225,6 @@ pub struct ExecCommandApprovalParams {
    pub command: Vec<String>,
    pub cwd: PathBuf,
    pub reason: Option<String>,
-    pub risk: Option<SandboxCommandAssessment>,
    pub parsed_cmd: Vec<ParsedCommand>,
 }

--- a/codex-rs/app-server-protocol/src/protocol/v2.rs
+++ b/codex-rs/app-server-protocol/src/protocol/v2.rs
@@ -2,15 +2,20 @@ use std::collections::HashMap;
 use std::path::PathBuf;

 use crate::protocol::common::AuthMode;
-use codex_protocol::ConversationId;
 use codex_protocol::account::PlanType;
-use codex_protocol::approvals::SandboxCommandAssessment as CoreSandboxCommandAssessment;
-use codex_protocol::config_types::ReasoningEffort;
+use codex_protocol::approvals::ExecPolicyAmendment as CoreExecPolicyAmendment;
+use codex_protocol::config_types::ForcedLoginMethod;
 use codex_protocol::config_types::ReasoningSummary;
+use codex_protocol::config_types::SandboxMode as CoreSandboxMode;
+use codex_protocol::config_types::Verbosity;
 use codex_protocol::items::AgentMessageContent as CoreAgentMessageContent;
 use codex_protocol::items::TurnItem as CoreTurnItem;
 use codex_protocol::models::ResponseItem;
+use codex_protocol::openai_models::ReasoningEffort;
 use codex_protocol::parse_command::ParsedCommand as CoreParsedCommand;
+use codex_protocol::plan_tool::PlanItemArg as CorePlanItemArg;
+use codex_protocol::plan_tool::StepStatus as CorePlanStepStatus;
+use codex_protocol::protocol::AskForApproval as CoreAskForApproval;
 use codex_protocol::protocol::CodexErrorInfo as CoreCodexErrorInfo;
 use codex_protocol::protocol::CreditsSnapshot as CoreCreditsSnapshot;
 use codex_protocol::protocol::RateLimitSnapshot as CoreRateLimitSnapshot;
@@ -20,6 +25,9 @@ use codex_protocol::protocol::TokenUsage as CoreTokenUsage;
 use codex_protocol::protocol::TokenUsageInfo as CoreTokenUsageInfo;
 use codex_protocol::user_input::UserInput as CoreUserInput;
 use mcp_types::ContentBlock as McpContentBlock;
+use mcp_types::Resource as McpResource;
+use mcp_types::ResourceTemplate as McpResourceTemplate;
+use mcp_types::Tool as McpTool;
 use schemars::JsonSchema;
 use serde::Deserialize;
 use serde::Serialize;
@@ -118,15 +126,81 @@ impl From<CoreCodexErrorInfo> for CodexErrorInfo {
    }
 }

+#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, JsonSchema, TS)]
+#[serde(rename_all = "kebab-case")]
+#[ts(rename_all = "kebab-case", export_to = "v2/")]
+pub enum AskForApproval {
+    #[serde(rename = "untrusted")]
+    #[ts(rename = "untrusted")]
+    UnlessTrusted,
+    OnFailure,
+    OnRequest,
+    Never,
+}
+
+impl AskForApproval {
+    pub fn to_core(self) -> CoreAskForApproval {
+        match self {
+            AskForApproval::UnlessTrusted => CoreAskForApproval::UnlessTrusted,
+            AskForApproval::OnFailure => CoreAskForApproval::OnFailure,
+            AskForApproval::OnRequest => CoreAskForApproval::OnRequest,
+            AskForApproval::Never => CoreAskForApproval::Never,
+        }
+    }
+}
+
+impl From<CoreAskForApproval> for AskForApproval {
+    fn from(value: CoreAskForApproval) -> Self {
+        match value {
+            CoreAskForApproval::UnlessTrusted => AskForApproval::UnlessTrusted,
+            CoreAskForApproval::OnFailure => AskForApproval::OnFailure,
+            CoreAskForApproval::OnRequest => AskForApproval::OnRequest,
+            CoreAskForApproval::Never => AskForApproval::Never,
+        }
+    }
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, JsonSchema, TS)]
+#[serde(rename_all = "kebab-case")]
+#[ts(rename_all = "kebab-case", export_to = "v2/")]
+pub enum SandboxMode {
+    ReadOnly,
+    WorkspaceWrite,
+    DangerFullAccess,
+}
+
+impl SandboxMode {
+    pub fn to_core(self) -> CoreSandboxMode {
+        match self {
+            SandboxMode::ReadOnly => CoreSandboxMode::ReadOnly,
+            SandboxMode::WorkspaceWrite => CoreSandboxMode::WorkspaceWrite,
+            SandboxMode::DangerFullAccess => CoreSandboxMode::DangerFullAccess,
+        }
+    }
+}
+
+impl From<CoreSandboxMode> for SandboxMode {
+    fn from(value: CoreSandboxMode) -> Self {
+        match value {
+            CoreSandboxMode::ReadOnly => SandboxMode::ReadOnly,
+            CoreSandboxMode::WorkspaceWrite => SandboxMode::WorkspaceWrite,
+            CoreSandboxMode::DangerFullAccess => SandboxMode::DangerFullAccess,
+        }
+    }
+}
+
 v2_enum_from_core!(
-    pub enum AskForApproval from codex_protocol::protocol::AskForApproval {
-        UnlessTrusted, OnFailure, OnRequest, Never
+    pub enum ReviewDelivery from codex_protocol::protocol::ReviewDelivery {
+        Inline, Detached
    }
 );

 v2_enum_from_core!(
-    pub enum SandboxMode from codex_protocol::config_types::SandboxMode {
-        ReadOnly, WorkspaceWrite, DangerFullAccess
+    pub enum McpAuthStatus from codex_protocol::protocol::McpAuthStatus {
+        Unsupported,
+        NotLoggedIn,
+        BearerToken,
+        OAuth
    }
 );

@@ -140,6 +214,72 @@ pub enum ConfigLayerName {
    User,
 }

+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, JsonSchema, TS)]
+#[serde(rename_all = "snake_case")]
+#[ts(export_to = "v2/")]
+pub struct SandboxWorkspaceWrite {
+    #[serde(default)]
+    pub writable_roots: Vec<PathBuf>,
+    #[serde(default)]
+    pub network_access: bool,
+    #[serde(default)]
+    pub exclude_tmpdir_env_var: bool,
+    #[serde(default)]
+    pub exclude_slash_tmp: bool,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "snake_case")]
+#[ts(export_to = "v2/")]
+pub struct ToolsV2 {
+    #[serde(alias = "web_search_request")]
+    pub web_search: Option<bool>,
+    pub view_image: Option<bool>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "snake_case")]
+#[ts(export_to = "v2/")]
+pub struct ProfileV2 {
+    pub model: Option<String>,
+    pub model_provider: Option<String>,
+    pub approval_policy: Option<AskForApproval>,
+    pub model_reasoning_effort: Option<ReasoningEffort>,
+    pub model_reasoning_summary: Option<ReasoningSummary>,
+    pub model_verbosity: Option<Verbosity>,
+    pub chatgpt_base_url: Option<String>,
+    #[serde(default, flatten)]
+    pub additional: HashMap<String, JsonValue>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "snake_case")]
+#[ts(export_to = "v2/")]
+pub struct Config {
+    pub model: Option<String>,
+    pub review_model: Option<String>,
+    pub model_context_window: Option<i64>,
+    pub model_auto_compact_token_limit: Option<i64>,
+    pub model_provider: Option<String>,
+    pub approval_policy: Option<AskForApproval>,
+    pub sandbox_mode: Option<SandboxMode>,
+    pub sandbox_workspace_write: Option<SandboxWorkspaceWrite>,
+    pub forced_chatgpt_workspace_id: Option<String>,
+    pub forced_login_method: Option<ForcedLoginMethod>,
+    pub tools: Option<ToolsV2>,
+    pub profile: Option<String>,
+    #[serde(default)]
+    pub profiles: HashMap<String, ProfileV2>,
+    pub instructions: Option<String>,
+    pub developer_instructions: Option<String>,
+    pub compact_prompt: Option<String>,
+    pub model_reasoning_effort: Option<ReasoningEffort>,
+    pub model_reasoning_summary: Option<ReasoningSummary>,
+    pub model_verbosity: Option<Verbosity>,
+    #[serde(default, flatten)]
+    pub additional: HashMap<String, JsonValue>,
+}
+
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
@@ -190,6 +330,8 @@ pub struct OverriddenMetadata {
 pub struct ConfigWriteResponse {
    pub status: WriteStatus,
    pub version: String,
+    /// Canonical path to the config file that was written.
+    pub file_path: String,
    pub overridden_metadata: Option<OverriddenMetadata>,
 }

@@ -216,7 +358,7 @@ pub struct ConfigReadParams {
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct ConfigReadResponse {
-    pub config: JsonValue,
+    pub config: Config,
    pub origins: HashMap<String, ConfigLayerMetadata>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub layers: Option<Vec<ConfigLayer>>,
@@ -226,10 +368,11 @@ pub struct ConfigReadResponse {
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct ConfigValueWriteParams {
-    pub file_path: String,
    pub key_path: String,
    pub value: JsonValue,
    pub merge_strategy: MergeStrategy,
+    /// Path to the config file to write; defaults to the user's `config.toml` when omitted.
+    pub file_path: Option<String>,
    pub expected_version: Option<String>,
 }

@@ -237,8 +380,9 @@ pub struct ConfigValueWriteParams {
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct ConfigBatchWriteParams {
-    pub file_path: String,
    pub edits: Vec<ConfigEdit>,
+    /// Path to the config file to write; defaults to the user's `config.toml` when omitted.
+    pub file_path: Option<String>,
    pub expected_version: Option<String>,
 }

@@ -251,19 +395,16 @@ pub struct ConfigEdit {
    pub merge_strategy: MergeStrategy,
 }

-v2_enum_from_core!(
-    pub enum CommandRiskLevel from codex_protocol::approvals::SandboxRiskLevel {
-        Low,
-        Medium,
-        High
-    }
-);
-
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub enum ApprovalDecision {
    Accept,
+    /// Approve and remember the approval for the session.
+    AcceptForSession,
+    AcceptWithExecpolicyAmendment {
+        execpolicy_amendment: ExecPolicyAmendment,
+    },
    Decline,
    Cancel,
 }
@@ -333,28 +474,23 @@ impl From<codex_protocol::protocol::SandboxPolicy> for SandboxPolicy {
    }
 }

-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
-#[serde(rename_all = "camelCase")]
-#[ts(export_to = "v2/")]
-pub struct SandboxCommandAssessment {
-    pub description: String,
-    pub risk_level: CommandRiskLevel,
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
+#[serde(transparent)]
+#[ts(type = "Array<string>", export_to = "v2/")]
+pub struct ExecPolicyAmendment {
+    pub command: Vec<String>,
 }

-impl SandboxCommandAssessment {
-    pub fn into_core(self) -> CoreSandboxCommandAssessment {
-        CoreSandboxCommandAssessment {
-            description: self.description,
-            risk_level: self.risk_level.to_core(),
-        }
+impl ExecPolicyAmendment {
+    pub fn into_core(self) -> CoreExecPolicyAmendment {
+        CoreExecPolicyAmendment::new(self.command)
    }
 }

-impl From<CoreSandboxCommandAssessment> for SandboxCommandAssessment {
-    fn from(value: CoreSandboxCommandAssessment) -> Self {
+impl From<CoreExecPolicyAmendment> for ExecPolicyAmendment {
+    fn from(value: CoreExecPolicyAmendment) -> Self {
        Self {
-            description: value.description,
-            risk_level: CommandRiskLevel::from(value.risk_level),
+            command: value.command().to_vec(),
        }
    }
 }
@@ -532,10 +668,21 @@ pub struct CancelLoginAccountParams {
    pub login_id: String,
 }

-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub enum CancelLoginAccountStatus {
+    Canceled,
+    NotFound,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
-pub struct CancelLoginAccountResponse {}
+pub struct CancelLoginAccountResponse {
+    pub status: CancelLoginAccountStatus,
+}

 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
@@ -607,13 +754,64 @@ pub struct ModelListResponse {
    pub next_cursor: Option<String>,
 }

+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct ListMcpServersParams {
+    /// Opaque pagination cursor returned by a previous call.
+    pub cursor: Option<String>,
+    /// Optional page size; defaults to a server-defined value.
+    pub limit: Option<u32>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct McpServer {
+    pub name: String,
+    pub tools: std::collections::HashMap<String, McpTool>,
+    pub resources: Vec<McpResource>,
+    pub resource_templates: Vec<McpResourceTemplate>,
+    pub auth_status: McpAuthStatus,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct ListMcpServersResponse {
+    pub data: Vec<McpServer>,
+    /// Opaque cursor to pass to the next call to continue after the last item.
+    /// If None, there are no more items to return.
+    pub next_cursor: Option<String>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct McpServerOauthLoginParams {
+    pub name: String,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    #[ts(optional)]
+    pub scopes: Option<Vec<String>>,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    #[ts(optional)]
+    pub timeout_secs: Option<i64>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct McpServerOauthLoginResponse {
+    pub authorization_url: String,
+}
+
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct FeedbackUploadParams {
    pub classification: String,
    pub reason: Option<String>,
-    pub conversation_id: Option<ConversationId>,
+    pub thread_id: Option<String>,
    pub include_logs: bool,
 }

@@ -624,6 +822,26 @@ pub struct FeedbackUploadResponse {
    pub thread_id: String,
 }

+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct CommandExecParams {
+    pub command: Vec<String>,
+    #[ts(type = "number | null")]
+    pub timeout_ms: Option<i64>,
+    pub cwd: Option<PathBuf>,
+    pub sandbox_policy: Option<SandboxPolicy>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct CommandExecResponse {
+    pub exit_code: i32,
+    pub stdout: String,
+    pub stderr: String,
+}
+
 // === Threads, Turns, and Items ===
 // Thread APIs
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Default, JsonSchema, TS)]
@@ -758,6 +976,7 @@ pub struct Thread {
    /// Model provider used for this thread (for example, 'openai').
    pub model_provider: String,
    /// Unix timestamp (in seconds) when the thread was created.
+    #[ts(type = "number")]
    pub created_at: i64,
    /// [UNSTABLE] Path to the thread on disk.
    pub path: PathBuf,
@@ -848,8 +1067,9 @@ pub struct Turn {
    /// For all other responses and notifications returning a Turn,
    /// the items field will be an empty list.
    pub items: Vec<ThreadItem>,
-    #[serde(flatten)]
    pub status: TurnStatus,
+    /// Only populated when the Turn's status is failed.
+    pub error: Option<TurnError>,
 }

 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS, Error)]
@@ -866,17 +1086,20 @@ pub struct TurnError {
 #[ts(export_to = "v2/")]
 pub struct ErrorNotification {
    pub error: TurnError,
+    // Set to true if the error is transient and the app-server process will automatically retry.
+    // If true, this will not interrupt a turn.
+    pub will_retry: bool,
    pub thread_id: String,
    pub turn_id: String,
 }

 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
-#[serde(tag = "status", rename_all = "camelCase")]
-#[ts(tag = "status", export_to = "v2/")]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
 pub enum TurnStatus {
    Completed,
    Interrupted,
-    Failed { error: TurnError },
+    Failed,
    InProgress,
 }

@@ -908,9 +1131,22 @@ pub struct ReviewStartParams {
    pub thread_id: String,
    pub target: ReviewTarget,

-    /// When true, also append the final review message to the original thread.
+    /// Where to run the review: inline (default) on the current thread or
+    /// detached on a new thread (returned in `reviewThreadId`).
    #[serde(default)]
-    pub append_to_original_thread: bool,
+    pub delivery: Option<ReviewDelivery>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct ReviewStartResponse {
+    pub turn: Turn,
+    /// Identifies the thread where the review runs.
+    ///
+    /// For inline reviews, this is the original thread id.
+    /// For detached reviews, this is the id of the new review thread.
+    pub review_thread_id: String,
 }

 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -1032,6 +1268,7 @@ pub enum ThreadItem {
        /// The command's exit code.
        exit_code: Option<i32>,
        /// The duration of the command execution in milliseconds.
+        #[ts(type = "number | null")]
        duration_ms: Option<i64>,
    },
    #[serde(rename_all = "camelCase")]
@@ -1051,19 +1288,22 @@ pub enum ThreadItem {
        arguments: JsonValue,
        result: Option<McpToolCallResult>,
        error: Option<McpToolCallError>,
+        /// The duration of the MCP tool call in milliseconds.
+        #[ts(type = "number | null")]
+        duration_ms: Option<i64>,
    },
    #[serde(rename_all = "camelCase")]
    #[ts(rename_all = "camelCase")]
    WebSearch { id: String, query: String },
    #[serde(rename_all = "camelCase")]
    #[ts(rename_all = "camelCase")]
-    TodoList { id: String, items: Vec<TodoItem> },
-    #[serde(rename_all = "camelCase")]
-    #[ts(rename_all = "camelCase")]
    ImageView { id: String, path: String },
    #[serde(rename_all = "camelCase")]
    #[ts(rename_all = "camelCase")]
-    CodeReview { id: String, review: String },
+    EnteredReviewMode { id: String, review: String },
+    #[serde(rename_all = "camelCase")]
+    #[ts(rename_all = "camelCase")]
+    ExitedReviewMode { id: String, review: String },
 }

 impl From<CoreTurnItem> for ThreadItem {
@@ -1159,15 +1399,6 @@ pub struct McpToolCallError {
    pub message: String,
 }

-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
-#[serde(rename_all = "camelCase")]
-#[ts(export_to = "v2/")]
-pub struct TodoItem {
-    pub id: String,
-    pub text: String,
-    pub completed: bool,
-}
-
 // === Server Notifications ===
 // Thread/Turn lifecycle notifications and item progress events
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -1208,10 +1439,57 @@ pub struct TurnCompletedNotification {
 /// Notification that the turn-level unified diff has changed.
 /// Contains the latest aggregated diff across all file changes in the turn.
 pub struct TurnDiffUpdatedNotification {
+    pub thread_id: String,
    pub turn_id: String,
    pub diff: String,
 }

+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct TurnPlanUpdatedNotification {
+    pub thread_id: String,
+    pub turn_id: String,
+    pub explanation: Option<String>,
+    pub plan: Vec<TurnPlanStep>,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, Eq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct TurnPlanStep {
+    pub step: String,
+    pub status: TurnPlanStepStatus,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, Copy, PartialEq, Eq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub enum TurnPlanStepStatus {
+    Pending,
+    InProgress,
+    Completed,
+}
+
+impl From<CorePlanItemArg> for TurnPlanStep {
+    fn from(value: CorePlanItemArg) -> Self {
+        Self {
+            step: value.step,
+            status: value.status.into(),
+        }
+    }
+}
+
+impl From<CorePlanStepStatus> for TurnPlanStepStatus {
+    fn from(value: CorePlanStepStatus) -> Self {
+        match value {
+            CorePlanStepStatus::Pending => Self::Pending,
+            CorePlanStepStatus::InProgress => Self::InProgress,
+            CorePlanStepStatus::Completed => Self::Completed,
+        }
+    }
+}
+
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
@@ -1235,6 +1513,8 @@ pub struct ItemCompletedNotification {
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct AgentMessageDeltaNotification {
+    pub thread_id: String,
+    pub turn_id: String,
    pub item_id: String,
    pub delta: String,
 }
@@ -1243,8 +1523,11 @@ pub struct AgentMessageDeltaNotification {
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct ReasoningSummaryTextDeltaNotification {
+    pub thread_id: String,
+    pub turn_id: String,
    pub item_id: String,
    pub delta: String,
+    #[ts(type = "number")]
    pub summary_index: i64,
 }

@@ -1252,7 +1535,10 @@ pub struct ReasoningSummaryTextDeltaNotification {
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct ReasoningSummaryPartAddedNotification {
+    pub thread_id: String,
+    pub turn_id: String,
    pub item_id: String,
+    #[ts(type = "number")]
    pub summary_index: i64,
 }

@@ -1260,15 +1546,41 @@ pub struct ReasoningSummaryPartAddedNotification {
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct ReasoningTextDeltaNotification {
+    pub thread_id: String,
+    pub turn_id: String,
    pub item_id: String,
    pub delta: String,
+    #[ts(type = "number")]
    pub content_index: i64,
 }

+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct TerminalInteractionNotification {
+    pub thread_id: String,
+    pub turn_id: String,
+    pub item_id: String,
+    pub process_id: String,
+    pub stdin: String,
+}
+
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct CommandExecutionOutputDeltaNotification {
+    pub thread_id: String,
+    pub turn_id: String,
+    pub item_id: String,
+    pub delta: String,
+}
+
+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct FileChangeOutputDeltaNotification {
+    pub thread_id: String,
+    pub turn_id: String,
    pub item_id: String,
    pub delta: String,
 }
@@ -1277,10 +1589,23 @@ pub struct CommandExecutionOutputDeltaNotification {
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
 pub struct McpToolCallProgressNotification {
+    pub thread_id: String,
+    pub turn_id: String,
    pub item_id: String,
    pub message: String,
 }

+#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
+#[serde(rename_all = "camelCase")]
+#[ts(export_to = "v2/")]
+pub struct McpServerOauthLoginCompletedNotification {
+    pub name: String,
+    pub success: bool,
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    #[ts(optional)]
+    pub error: Option<String>,
+}
+
 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
 #[serde(rename_all = "camelCase")]
 #[ts(export_to = "v2/")]
@@ -1307,17 +1632,8 @@ pub struct CommandExecutionRequestApprovalParams {
    pub item_id: String,
    /// Optional explanatory reason (e.g. request for network access).
    pub reason: Option<String>,
-    /// Optional model-provided risk assessment describing the blocked command.
-    pub risk: Option<SandboxCommandAssessment>,
-}
-
-#[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
-#[serde(rename_all = "camelCase")]
-#[ts(export_to = "v2/")]
-pub struct CommandExecutionRequestAcceptSettings {
-    /// If true, automatically approve this command for the duration of the session.
-    #[serde(default)]
-    pub for_session: bool,
+    /// Optional proposed execpolicy amendment to allow similar commands without prompting.
+    pub proposed_execpolicy_amendment: Option<ExecPolicyAmendment>,
 }

 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -1325,10 +1641,6 @@ pub struct CommandExecutionRequestAcceptSettings {
 #[ts(export_to = "v2/")]
 pub struct CommandExecutionRequestApprovalResponse {
    pub decision: ApprovalDecision,
-    /// Optional approval settings for when the decision is `accept`.
-    /// Ignored if the decision is `decline` or `cancel`.
-    #[serde(default)]
-    pub accept_settings: Option<CommandExecutionRequestAcceptSettings>,
 }

 #[derive(Serialize, Deserialize, Debug, Clone, PartialEq, JsonSchema, TS)]
@@ -1365,6 +1677,7 @@ pub struct RateLimitSnapshot {
    pub primary: Option<RateLimitWindow>,
    pub secondary: Option<RateLimitWindow>,
    pub credits: Option<CreditsSnapshot>,
+    pub plan_type: Option<PlanType>,
 }

 impl From<CoreRateLimitSnapshot> for RateLimitSnapshot {
@@ -1373,6 +1686,7 @@ impl From<CoreRateLimitSnapshot> for RateLimitSnapshot {
            primary: value.primary.map(RateLimitWindow::from),
            secondary: value.secondary.map(RateLimitWindow::from),
            credits: value.credits.map(CreditsSnapshot::from),
+            plan_type: value.plan_type,
        }
    }
 }
@@ -1382,7 +1696,9 @@ impl From<CoreRateLimitSnapshot> for RateLimitSnapshot {
 #[ts(export_to = "v2/")]
 pub struct RateLimitWindow {
    pub used_percent: i32,
+    #[ts(type = "number | null")]
    pub window_duration_mins: Option<i64>,
+    #[ts(type = "number | null")]
    pub resets_at: Option<i64>,
 }

--- a/codex-rs/app-server-test-client/src/main.rs
+++ b/codex-rs/app-server-test-client/src/main.rs
@@ -21,7 +21,6 @@ use codex_app_server_protocol::ApprovalDecision;
 use codex_app_server_protocol::AskForApproval;
 use codex_app_server_protocol::ClientInfo;
 use codex_app_server_protocol::ClientRequest;
-use codex_app_server_protocol::CommandExecutionRequestAcceptSettings;
 use codex_app_server_protocol::CommandExecutionRequestApprovalParams;
 use codex_app_server_protocol::CommandExecutionRequestApprovalResponse;
 use codex_app_server_protocol::FileChangeRequestApprovalParams;
@@ -554,6 +553,10 @@ impl CodexClient {
                    print!("{}", delta.delta);
                    std::io::stdout().flush().ok();
                }
+                ServerNotification::TerminalInteraction(delta) => {
+                    println!("[stdin sent: {}]", delta.stdin);
+                    std::io::stdout().flush().ok();
+                }
                ServerNotification::ItemStarted(payload) => {
                    println!("\n< item started: {:?}", payload.item);
                }
@@ -563,7 +566,9 @@ impl CodexClient {
                ServerNotification::TurnCompleted(payload) => {
                    if payload.turn.id == turn_id {
                        println!("\n< turn/completed notification: {:?}", payload.turn.status);
-                        if let TurnStatus::Failed { error } = &payload.turn.status {
+                        if payload.turn.status == TurnStatus::Failed
+                            && let Some(error) = payload.turn.error
+                        {
                            println!("[turn error] {}", error.message);
                        }
                        break;
@@ -751,7 +756,7 @@ impl CodexClient {
            turn_id,
            item_id,
            reason,
-            risk,
+            proposed_execpolicy_amendment,
        } = params;

        println!(
@@ -760,13 +765,12 @@ impl CodexClient {
        if let Some(reason) = reason.as_deref() {
            println!("< reason: {reason}");
        }
-        if let Some(risk) = risk.as_ref() {
-            println!("< risk assessment: {risk:?}");
+        if let Some(execpolicy_amendment) = proposed_execpolicy_amendment.as_ref() {
+            println!("< proposed execpolicy amendment: {execpolicy_amendment:?}");
        }

        let response = CommandExecutionRequestApprovalResponse {
            decision: ApprovalDecision::Accept,
-            accept_settings: Some(CommandExecutionRequestAcceptSettings { for_session: false }),
        };
        self.send_server_request_response(request_id, &response)?;
        println!("< approved commandExecution request for item {item_id}");
--- a/codex-rs/app-server/Cargo.toml
+++ b/codex-rs/app-server/Cargo.toml
@@ -26,13 +26,16 @@ codex-login = { workspace = true }
 codex-protocol = { workspace = true }
 codex-app-server-protocol = { workspace = true }
 codex-feedback = { workspace = true }
+codex-rmcp-client = { workspace = true }
 codex-utils-json-to-toml = { workspace = true }
 chrono = { workspace = true }
 serde = { workspace = true, features = ["derive"] }
 serde_json = { workspace = true }
 sha2 = { workspace = true }
+mcp-types = { workspace = true }
 tempfile = { workspace = true }
 toml = { workspace = true }
+toml_edit = { workspace = true }
 tokio = { workspace = true, features = [
    "io-std",
    "macros",
--- a/codex-rs/app-server/README.md
+++ b/codex-rs/app-server/README.md
@@ -1,15 +1,15 @@
 # codex-app-server

-`codex app-server` is the interface Codex uses to power rich interfaces such as the [Codex VS Code extension](https://marketplace.visualstudio.com/items?itemName=openai.chatgpt). The message schema is currently unstable, but those who wish to build experimental UIs on top of Codex may find it valuable.
+`codex app-server` is the interface Codex uses to power rich interfaces such as the [Codex VS Code extension](https://marketplace.visualstudio.com/items?itemName=openai.chatgpt).

 ## Table of Contents
 - [Protocol](#protocol)
 - [Message Schema](#message-schema)
+- [Core Primitives](#core-primitives)
 - [Lifecycle Overview](#lifecycle-overview)
 - [Initialization](#initialization)
- [Core primitives](#core-primitives)
- [Thread & turn endpoints](#thread--turn-endpoints)
- [Events (work-in-progress)](#events-work-in-progress)
+- [API Overview](#api-overview)
+- [Events](#events)
 - [Auth endpoints](#auth-endpoints)

 ## Protocol
@@ -25,6 +25,15 @@ codex app-server generate-ts --out DIR
 codex app-server generate-json-schema --out DIR
 ```

+## Core Primitives
+
+The API exposes three top level primitives representing an interaction between a user and Codex:
+- **Thread**: A conversation between a user and the Codex agent. Each thread contains multiple turns.
+- **Turn**: One turn of the conversation, typically starting with a user message and finishing with an agent message. Each turn contains multiple items.
+- **Item**: Represents user inputs and agent outputs as part of the turn, persisted and used as the context for future conversations. Example items include user message, agent reasoning, agent message, shell command, file edit, etc.
+
+Use the thread APIs to create, list, or archive conversations. Drive a conversation with turn APIs and stream progress via turn notifications.
+
 ## Lifecycle Overview

 - Initialize once: Immediately after launching the codex app-server process, send an `initialize` request with your client metadata, then emit an `initialized` notification. Any other request before this handshake gets rejected.
@@ -37,37 +46,34 @@ codex app-server generate-json-schema --out DIR

 Clients must send a single `initialize` request before invoking any other method, then acknowledge with an `initialized` notification. The server returns the user agent string it will present to upstream services; subsequent requests issued before initialization receive a `"Not initialized"` error, and repeated `initialize` calls receive an `"Already initialized"` error.

-Example:
+Applications building on top of `codex app-server` should identify themselves via the `clientInfo` parameter.

+Example (from OpenAI's official VSCode extension):
 ```json
 { "method": "initialize", "id": 0, "params": {
    "clientInfo": { "name": "codex-vscode", "title": "Codex VS Code Extension", "version": "0.1.0" }
 } }
-{ "id": 0, "result": { "userAgent": "codex-app-server/0.1.0 codex-vscode/0.1.0" } }
-{ "method": "initialized" }
 ```

-## Core primitives
-
-We have 3 top level primitives:
- Thread - a conversation between the Codex agent and a user. Each thread contains multiple turns.
- Turn - one turn of the conversation, typically starting with a user message and finishing with an agent message. Each turn contains multiple items.
- Item - represents user inputs and agent outputs as part of the turn, persisted and used as the context for future conversations.
-
-## Thread & turn endpoints
-
-The JSON-RPC API exposes dedicated methods for managing Codex conversations. Threads store long-lived conversation metadata, and turns store the per-message exchange (input → Codex output, including streamed items). Use the thread APIs to create, list, or archive sessions, then drive the conversation with turn APIs and notifications.
-
-### Quick reference
+## API Overview
 - `thread/start` — create a new thread; emits `thread/started` and auto-subscribes you to turn/item events for that thread.
 - `thread/resume` — reopen an existing thread by id so subsequent `turn/start` calls append to it.
 - `thread/list` — page through stored rollouts; supports cursor-based pagination and optional `modelProviders` filtering.
 - `thread/archive` — move a thread’s rollout file into the archived directory; returns `{}` on success.
 - `turn/start` — add user input to a thread and begin Codex generation; responds with the initial `turn` object and streams `turn/started`, `item/*`, and `turn/completed` notifications.
 - `turn/interrupt` — request cancellation of an in-flight turn by `(thread_id, turn_id)`; success is an empty `{}` response and the turn finishes with `status: "interrupted"`.
- `review/start` — kick off Codex’s automated reviewer for a thread; responds like `turn/start` and emits a `item/completed` notification with a `codeReview` item when results are ready.
+- `review/start` — kick off Codex’s automated reviewer for a thread; responds like `turn/start` and emits `item/started`/`item/completed` notifications with `enteredReviewMode` and `exitedReviewMode` items, plus a final assistant `agentMessage` containing the review.
+- `command/exec` — run a single command under the server sandbox without starting a thread/turn (handy for utilities and validation).
+- `model/list` — list available models (with reasoning effort options).
+- `mcpServer/oauth/login` — start an OAuth login for a configured MCP server; returns an `authorization_url` and later emits `mcpServer/oauthLogin/completed` once the browser flow finishes.
+- `mcpServers/list` — enumerate configured MCP servers with their tools, resources, resource templates, and auth status; supports cursor+limit pagination.
+- `feedback/upload` — submit a feedback report (classification + optional reason/logs and conversation_id); returns the tracking thread id.
+- `command/exec` — run a single command under the server sandbox without starting a thread/turn (handy for utilities and validation).
+- `config/read` — fetch the effective config on disk after resolving config layering.
+- `config/value/write` — write a single config key/value to the user's config.toml on disk.
+- `config/batchWrite` — apply multiple config edits atomically to the user's config.toml on disk.

-### 1) Start or resume a thread
+### Example: Start or resume a thread

 Start a fresh thread when you need a new Codex conversation.

@@ -98,7 +104,7 @@ To continue a stored session, call `thread/resume` with the `thread.id` you prev
 { "id": 11, "result": { "thread": { "id": "thr_123", … } } }
 ```

-### 2) List threads (pagination & filters)
+### Example: List threads (with pagination & filters)

 `thread/list` lets you render a history UI. Pass any combination of:
 - `cursor` — opaque string from a prior response; omit for the first page.
@@ -123,7 +129,7 @@ Example:

 When `nextCursor` is `null`, you’ve reached the final page.

-### 3) Archive a thread
+### Example: Archive a thread

 Use `thread/archive` to move the persisted rollout (stored as a JSONL file on disk) into the archived sessions directory.

@@ -134,7 +140,7 @@ Use `thread/archive` to move the persisted rollout (stored as a JSONL file on di

 An archived thread will not appear in future calls to `thread/list`.

-### 4) Start a turn (send user input)
+### Example: Start a turn (send user input)

 Turns attach user input (text or images) to a thread and trigger Codex generation. The `input` field is a list of discriminated unions:

@@ -168,7 +174,7 @@ You can optionally specify config overrides on the new turn. If specified, these
 } } }
 ```

-### 5) Interrupt an active turn
+### Example: Interrupt an active turn

 You can cancel a running Turn with `turn/interrupt`.

@@ -182,7 +188,7 @@ You can cancel a running Turn with `turn/interrupt`.

 The server requests cancellations for running subprocesses, then emits a `turn/completed` event with `status: "interrupted"`. Rely on the `turn/completed` to know when Codex-side cleanup is done.

-### 6) Request a code review
+### Example: Request a code review

 Use `review/start` to run Codex’s reviewer on the currently checked-out project. The request takes the thread id plus a `target` describing what should be reviewed:

@@ -190,51 +196,77 @@ Use `review/start` to run Codex’s reviewer on the currently checked-out projec
 - `{"type":"baseBranch","branch":"main"}` — diff against the provided branch’s upstream (see prompt for the exact `git merge-base`/`git diff` instructions Codex will run).
 - `{"type":"commit","sha":"abc1234","title":"Optional subject"}` — review a specific commit.
 - `{"type":"custom","instructions":"Free-form reviewer instructions"}` — fallback prompt equivalent to the legacy manual review request.
- `appendToOriginalThread` (bool, default `false`) — when `true`, Codex also records a final assistant-style message with the review summary in the original thread. When `false`, only the `codeReview` item is emitted for the review run and no extra message is added to the original thread.
+- `delivery` (`"inline"` or `"detached"`, default `"inline"`) — where the review runs:
+  - `"inline"`: run the review as a new turn on the existing thread. The response’s `reviewThreadId` equals the original `threadId`, and no new `thread/started` notification is emitted.
+  - `"detached"`: fork a new review thread from the parent conversation and run the review there. The response’s `reviewThreadId` is the id of this new review thread, and the server emits a `thread/started` notification for it before streaming review items.

 Example request/response:

 ```json
 { "method": "review/start", "id": 40, "params": {
    "threadId": "thr_123",
-    "appendToOriginalThread": true,
+    "delivery": "inline",
    "target": { "type": "commit", "sha": "1234567deadbeef", "title": "Polish tui colors" }
 } }
-{ "id": 40, "result": { "turn": {
-    "id": "turn_900",
-    "status": "inProgress",
-    "items": [
-        { "type": "userMessage", "id": "turn_900", "content": [ { "type": "text", "text": "Review commit 1234567: Polish tui colors" } ] }
-    ],
-    "error": null
-} } }
+{ "id": 40, "result": {
+    "turn": {
+        "id": "turn_900",
+        "status": "inProgress",
+        "items": [
+            { "type": "userMessage", "id": "turn_900", "content": [ { "type": "text", "text": "Review commit 1234567: Polish tui colors" } ] }
+        ],
+        "error": null
+    },
+    "reviewThreadId": "thr_123"
+} }
 ```

+For a detached review, use `"delivery": "detached"`. The response is the same shape, but `reviewThreadId` will be the id of the new review thread (different from the original `threadId`). The server also emits a `thread/started` notification for that new thread before streaming the review turn.
+
 Codex streams the usual `turn/started` notification followed by an `item/started`
-with the same `codeReview` item id so clients can show progress:
+with an `enteredReviewMode` item so clients can show progress:

 ```json
 { "method": "item/started", "params": { "item": {
-    "type": "codeReview",
+    "type": "enteredReviewMode",
    "id": "turn_900",
    "review": "current changes"
 } } }
 ```

-When the reviewer finishes, the server emits `item/completed` containing the same
-`codeReview` item with the final review text:
+When the reviewer finishes, the server emits `item/started` and `item/completed`
+containing an `exitedReviewMode` item with the final review text:

 ```json
 { "method": "item/completed", "params": { "item": {
-    "type": "codeReview",
+    "type": "exitedReviewMode",
    "id": "turn_900",
    "review": "Looks solid overall...\n\n- Prefer Stylize helpers — app.rs:10-20\n  ..."
 } } }
 ```

-The `review` string is plain text that already bundles the overall explanation plus a bullet list for each structured finding (matching `ThreadItem::CodeReview` in the generated schema). Use this notification to render the reviewer output in your client.
+The `review` string is plain text that already bundles the overall explanation plus a bullet list for each structured finding (matching `ThreadItem::ExitedReviewMode` in the generated schema). Use this notification to render the reviewer output in your client.

-## Events (work-in-progress)
+### Example: One-off command execution
+
+Run a standalone command (argv vector) in the server’s sandbox without creating a thread or turn:
+
+```json
+{ "method": "command/exec", "id": 32, "params": {
+    "command": ["ls", "-la"],
+    "cwd": "/Users/me/project",                    // optional; defaults to server cwd
+    "sandboxPolicy": { "type": "workspaceWrite" }, // optional; defaults to user config
+    "timeoutMs": 10000                             // optional; ms timeout; defaults to server timeout
+} }
+{ "id": 32, "result": { "exitCode": 0, "stdout": "...", "stderr": "" } }
+```
+
+Notes:
+- Empty `command` arrays are rejected.
+- `sandboxPolicy` accepts the same shape used by `turn/start` (e.g., `dangerFullAccess`, `readOnly`, `workspaceWrite` with flags).
+- When omitted, `timeoutMs` falls back to the server default.
+
+## Events

 Event notifications are the server-initiated event stream for thread lifecycles, turn lifecycles, and the items within them. After you start or resume a thread, keep reading stdout for `thread/started`, `turn/*`, and `item/*` notifications.

@@ -244,10 +276,12 @@ The app-server streams JSON-RPC notifications while a turn is running. Each turn

 - `turn/started` — `{ turn }` with the turn id, empty `items`, and `status: "inProgress"`.
 - `turn/completed` — `{ turn }` where `turn.status` is `completed`, `interrupted`, or `failed`; failures carry `{ error: { message, codexErrorInfo? } }`.
+- `turn/diff/updated` — `{ threadId, turnId, diff }` represents the up-to-date snapshot of the turn-level unified diff, emitted after every FileChange item. `diff` is the latest aggregated unified diff across every file change in the turn. UIs can render this to show the full "what changed" view without stitching individual `fileChange` items.
+- `turn/plan/updated` — `{ turnId, explanation?, plan }` whenever the agent shares or changes its plan; each `plan` entry is `{ step, status }` with `status` in `pending`, `inProgress`, or `completed`.

 Today both notifications carry an empty `items` array even when item events were streamed; rely on `item/*` notifications for the canonical item list until this is fixed.

-#### Thread items
+#### Items

 `ThreadItem` is the tagged union carried in turn responses and `item/*` notifications. Currently we support events for the following items:
 - `userMessage` — `{id, content}` where `content` is a list of user inputs (`text`, `image`, or `localImage`).
@@ -257,6 +291,9 @@ Today both notifications carry an empty `items` array even when item events were
 - `fileChange` — `{id, changes, status}` describing proposed edits; `changes` list `{path, kind, diff}` and `status` is `inProgress`, `completed`, `failed`, or `declined`.
 - `mcpToolCall` — `{id, server, tool, status, arguments, result?, error?}` describing MCP calls; `status` is `inProgress`, `completed`, or `failed`.
 - `webSearch` — `{id, query}` for a web search request issued by the agent.
+- `imageView` — `{id, path}` emitted when the agent invokes the image viewer tool.
+- `enteredReviewMode` — `{id, review}` sent when the reviewer starts; `review` is a short user-facing label such as `"current changes"` or the requested target description.
+- `exitedReviewMode` — `{id, review}` emitted when the reviewer finishes; `review` is the full plain-text review (usually, overall notes plus bullet point findings).
 - `compacted` - `{threadId, turnId}` when codex compacts the conversation history. This can happen automatically.

 All items emit two shared lifecycle events:
@@ -274,7 +311,7 @@ There are additional item-specific events:
 - `item/commandExecution/outputDelta` — streams stdout/stderr for the command; append deltas in order to render live output alongside `aggregatedOutput` in the final item.
 Final `commandExecution` items include parsed `commandActions`, `status`, `exitCode`, and `durationMs` so the UI can summarize what ran and whether it succeeded.
 #### fileChange
-`fileChange` items contain a `changes` list with `{path, kind, diff}` entries (`kind` is `add`, `delete`, or `update` with an optional `movePath`). The `status` tracks whether apply succeeded (`completed`), failed, or was `declined`.
+- `item/fileChange/outputDelta` - contains the tool call response of the underlying `apply_patch` tool call.

 ### Errors
 `error` event is emitted whenever the server hits an error mid-turn (for example, upstream model errors or quota limits). Carries the same `{ error: { message, codexErrorInfo? } }` payload as `turn.status: "failed"` and may precede that terminal notification.
@@ -323,7 +360,7 @@ UI guidance for IDEs: surface an approval dialog as soon as the request arrives.

 The JSON-RPC auth/account surface exposes request/response methods plus server-initiated notifications (no `id`). Use these to determine auth state, start or cancel logins, logout, and inspect ChatGPT rate limits.

-### Quick reference
+### API Overview
 - `account/read` — fetch current account info; optionally refresh tokens.
 - `account/login/start` — begin login (`apiKey` or `chatgpt`).
 - `account/login/completed` (notify) — emitted when a login attempt finishes (success or error).
@@ -331,6 +368,8 @@ The JSON-RPC auth/account surface exposes request/response methods plus server-i
 - `account/logout` — sign out; triggers `account/updated`.
 - `account/updated` (notify) — emitted whenever auth mode changes (`authMode`: `apikey`, `chatgpt`, or `null`).
 - `account/rateLimits/read` — fetch ChatGPT rate limits; updates arrive via `account/rateLimits/updated` (notify).
+- `account/rateLimits/updated` (notify) — emitted whenever a user's ChatGPT rate limits change.
+- `mcpServer/oauthLogin/completed` (notify) — emitted after a `mcpServer/oauth/login` flow finishes for a server; payload includes `{ name, success, error? }`.

 ### 1) Check auth state

@@ -408,9 +447,3 @@ Field notes:
 - `usedPercent` is current usage within the OpenAI quota window.
 - `windowDurationMins` is the quota window length.
 - `resetsAt` is a Unix timestamp (seconds) for the next reset.
-
-### Dev notes
-
- `codex app-server generate-ts --out <dir>` emits v2 types under `v2/`.
- `codex app-server generate-json-schema --out <dir>` outputs `codex_app_server_protocol.schemas.json`.
- See [“Authentication and authorization” in the config docs](../../docs/config.md#authentication-and-authorization) for configuration knobs.
--- a/codex-rs/app-server/src/bespoke_event_handling.rs
+++ b/codex-rs/app-server/src/bespoke_event_handling.rs
@@ -18,6 +18,8 @@ use codex_app_server_protocol::ContextCompactedNotification;
 use codex_app_server_protocol::ErrorNotification;
 use codex_app_server_protocol::ExecCommandApprovalParams;
 use codex_app_server_protocol::ExecCommandApprovalResponse;
+use codex_app_server_protocol::ExecPolicyAmendment as V2ExecPolicyAmendment;
+use codex_app_server_protocol::FileChangeOutputDeltaNotification;
 use codex_app_server_protocol::FileChangeRequestApprovalParams;
 use codex_app_server_protocol::FileChangeRequestApprovalResponse;
 use codex_app_server_protocol::FileUpdateChange;
@@ -32,9 +34,9 @@ use codex_app_server_protocol::PatchChangeKind as V2PatchChangeKind;
 use codex_app_server_protocol::ReasoningSummaryPartAddedNotification;
 use codex_app_server_protocol::ReasoningSummaryTextDeltaNotification;
 use codex_app_server_protocol::ReasoningTextDeltaNotification;
-use codex_app_server_protocol::SandboxCommandAssessment as V2SandboxCommandAssessment;
 use codex_app_server_protocol::ServerNotification;
 use codex_app_server_protocol::ServerRequestPayload;
+use codex_app_server_protocol::TerminalInteractionNotification;
 use codex_app_server_protocol::ThreadItem;
 use codex_app_server_protocol::ThreadTokenUsage;
 use codex_app_server_protocol::ThreadTokenUsageUpdatedNotification;
@@ -43,6 +45,8 @@ use codex_app_server_protocol::TurnCompletedNotification;
 use codex_app_server_protocol::TurnDiffUpdatedNotification;
 use codex_app_server_protocol::TurnError;
 use codex_app_server_protocol::TurnInterruptResponse;
+use codex_app_server_protocol::TurnPlanStep;
+use codex_app_server_protocol::TurnPlanUpdatedNotification;
 use codex_app_server_protocol::TurnStatus;
 use codex_core::CodexConversation;
 use codex_core::parse_command::shlex_join;
@@ -59,7 +63,9 @@ use codex_core::protocol::ReviewDecision;
 use codex_core::protocol::TokenCountEvent;
 use codex_core::protocol::TurnDiffEvent;
 use codex_core::review_format::format_review_findings_block;
+use codex_core::review_prompts;
 use codex_protocol::ConversationId;
+use codex_protocol::plan_tool::UpdatePlanArgs;
 use codex_protocol::protocol::ReviewOutputEvent;
 use std::collections::HashMap;
 use std::convert::TryFrom;
@@ -173,7 +179,7 @@ pub(crate) async fn apply_bespoke_event_handling(
            command,
            cwd,
            reason,
-            risk,
+            proposed_execpolicy_amendment,
            parsed_cmd,
        }) => match api_version {
            ApiVersion::V1 => {
@@ -183,7 +189,6 @@ pub(crate) async fn apply_bespoke_event_handling(
                    command,
                    cwd,
                    reason,
-                    risk,
                    parsed_cmd,
                };
                let rx = outgoing
@@ -201,6 +206,8 @@ pub(crate) async fn apply_bespoke_event_handling(
                    .map(V2ParsedCommand::from)
                    .collect::<Vec<_>>();
                let command_string = shlex_join(&command);
+                let proposed_execpolicy_amendment_v2 =
+                    proposed_execpolicy_amendment.map(V2ExecPolicyAmendment::from);

                let params = CommandExecutionRequestApprovalParams {
                    thread_id: conversation_id.to_string(),
@@ -209,7 +216,7 @@ pub(crate) async fn apply_bespoke_event_handling(
                    // and emit the corresponding EventMsg, we repurpose the call_id as the item_id.
                    item_id: item_id.clone(),
                    reason,
-                    risk: risk.map(V2SandboxCommandAssessment::from),
+                    proposed_execpolicy_amendment: proposed_execpolicy_amendment_v2,
                };
                let rx = outgoing
                    .send_request(ServerRequestPayload::CommandExecutionRequestApproval(
@@ -257,6 +264,8 @@ pub(crate) async fn apply_bespoke_event_handling(
        }
        EventMsg::AgentMessageContentDelta(event) => {
            let notification = AgentMessageDeltaNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
                item_id: event.item_id,
                delta: event.delta,
            };
@@ -275,6 +284,8 @@ pub(crate) async fn apply_bespoke_event_handling(
        }
        EventMsg::ReasoningContentDelta(event) => {
            let notification = ReasoningSummaryTextDeltaNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
                item_id: event.item_id,
                delta: event.delta,
                summary_index: event.summary_index,
@@ -287,6 +298,8 @@ pub(crate) async fn apply_bespoke_event_handling(
        }
        EventMsg::ReasoningRawContentDelta(event) => {
            let notification = ReasoningTextDeltaNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
                item_id: event.item_id,
                delta: event.delta,
                content_index: event.content_index,
@@ -297,6 +310,8 @@ pub(crate) async fn apply_bespoke_event_handling(
        }
        EventMsg::AgentReasoningSectionBreak(event) => {
            let notification = ReasoningSummaryPartAddedNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
                item_id: event.item_id,
                summary_index: event.summary_index,
            };
@@ -319,6 +334,7 @@ pub(crate) async fn apply_bespoke_event_handling(
            outgoing
                .send_server_notification(ServerNotification::Error(ErrorNotification {
                    error: turn_error,
+                    will_retry: false,
                    thread_id: conversation_id.to_string(),
                    turn_id: event_turn_id.clone(),
                }))
@@ -334,22 +350,57 @@ pub(crate) async fn apply_bespoke_event_handling(
            outgoing
                .send_server_notification(ServerNotification::Error(ErrorNotification {
                    error: turn_error,
+                    will_retry: true,
                    thread_id: conversation_id.to_string(),
                    turn_id: event_turn_id.clone(),
                }))
                .await;
        }
-        EventMsg::EnteredReviewMode(review_request) => {
-            let notification = ItemStartedNotification {
+        EventMsg::ViewImageToolCall(view_image_event) => {
+            let item = ThreadItem::ImageView {
+                id: view_image_event.call_id.clone(),
+                path: view_image_event.path.to_string_lossy().into_owned(),
+            };
+            let started = ItemStartedNotification {
                thread_id: conversation_id.to_string(),
                turn_id: event_turn_id.clone(),
-                item: ThreadItem::CodeReview {
-                    id: event_turn_id.clone(),
-                    review: review_request.user_facing_hint,
-                },
+                item: item.clone(),
            };
            outgoing
-                .send_server_notification(ServerNotification::ItemStarted(notification))
+                .send_server_notification(ServerNotification::ItemStarted(started))
+                .await;
+            let completed = ItemCompletedNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
+                item,
+            };
+            outgoing
+                .send_server_notification(ServerNotification::ItemCompleted(completed))
+                .await;
+        }
+        EventMsg::EnteredReviewMode(review_request) => {
+            let review = review_request
+                .user_facing_hint
+                .unwrap_or_else(|| review_prompts::user_facing_hint(&review_request.target));
+            let item = ThreadItem::EnteredReviewMode {
+                id: event_turn_id.clone(),
+                review,
+            };
+            let started = ItemStartedNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
+                item: item.clone(),
+            };
+            outgoing
+                .send_server_notification(ServerNotification::ItemStarted(started))
+                .await;
+            let completed = ItemCompletedNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
+                item,
+            };
+            outgoing
+                .send_server_notification(ServerNotification::ItemCompleted(completed))
                .await;
        }
        EventMsg::ItemStarted(item_started_event) => {
@@ -375,21 +426,29 @@ pub(crate) async fn apply_bespoke_event_handling(
                .await;
        }
        EventMsg::ExitedReviewMode(review_event) => {
-            let review_text = match review_event.review_output {
+            let review = match review_event.review_output {
                Some(output) => render_review_output_text(&output),
                None => REVIEW_FALLBACK_MESSAGE.to_string(),
            };
-            let review_item_id = event_turn_id.clone();
-            let notification = ItemCompletedNotification {
+            let item = ThreadItem::ExitedReviewMode {
+                id: event_turn_id.clone(),
+                review,
+            };
+            let started = ItemStartedNotification {
                thread_id: conversation_id.to_string(),
                turn_id: event_turn_id.clone(),
-                item: ThreadItem::CodeReview {
-                    id: review_item_id,
-                    review: review_text,
-                },
+                item: item.clone(),
            };
            outgoing
-                .send_server_notification(ServerNotification::ItemCompleted(notification))
+                .send_server_notification(ServerNotification::ItemStarted(started))
+                .await;
+            let completed = ItemCompletedNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
+                item,
+            };
+            outgoing
+                .send_server_notification(ServerNotification::ItemCompleted(completed))
                .await;
        }
        EventMsg::PatchApplyBegin(patch_begin_event) => {
@@ -472,14 +531,57 @@ pub(crate) async fn apply_bespoke_event_handling(
                .await;
        }
        EventMsg::ExecCommandOutputDelta(exec_command_output_delta_event) => {
-            let notification = CommandExecutionOutputDeltaNotification {
-                item_id: exec_command_output_delta_event.call_id.clone(),
-                delta: String::from_utf8_lossy(&exec_command_output_delta_event.chunk).to_string(),
+            let item_id = exec_command_output_delta_event.call_id.clone();
+            let delta = String::from_utf8_lossy(&exec_command_output_delta_event.chunk).to_string();
+            // The underlying EventMsg::ExecCommandOutputDelta is used for shell, unified_exec,
+            // and apply_patch tool calls. We represent apply_patch with the FileChange item, and
+            // everything else with the CommandExecution item.
+            //
+            // We need to detect which item type it is so we can emit the right notification.
+            // We already have state tracking FileChange items on item/started, so let's use that.
+            let is_file_change = {
+                let map = turn_summary_store.lock().await;
+                map.get(&conversation_id)
+                    .is_some_and(|summary| summary.file_change_started.contains(&item_id))
+            };
+            if is_file_change {
+                let notification = FileChangeOutputDeltaNotification {
+                    thread_id: conversation_id.to_string(),
+                    turn_id: event_turn_id.clone(),
+                    item_id,
+                    delta,
+                };
+                outgoing
+                    .send_server_notification(ServerNotification::FileChangeOutputDelta(
+                        notification,
+                    ))
+                    .await;
+            } else {
+                let notification = CommandExecutionOutputDeltaNotification {
+                    thread_id: conversation_id.to_string(),
+                    turn_id: event_turn_id.clone(),
+                    item_id,
+                    delta,
+                };
+                outgoing
+                    .send_server_notification(ServerNotification::CommandExecutionOutputDelta(
+                        notification,
+                    ))
+                    .await;
+            }
+        }
+        EventMsg::TerminalInteraction(terminal_event) => {
+            let item_id = terminal_event.call_id.clone();
+
+            let notification = TerminalInteractionNotification {
+                thread_id: conversation_id.to_string(),
+                turn_id: event_turn_id.clone(),
+                item_id,
+                process_id: terminal_event.process_id,
+                stdin: terminal_event.stdin,
            };
            outgoing
-                .send_server_notification(ServerNotification::CommandExecutionOutputDelta(
-                    notification,
-                ))
+                .send_server_notification(ServerNotification::TerminalInteraction(notification))
                .await;
        }
        EventMsg::ExecCommandEnd(exec_command_end_event) => {
@@ -567,6 +669,7 @@ pub(crate) async fn apply_bespoke_event_handling(
        }
        EventMsg::TurnDiff(turn_diff_event) => {
            handle_turn_diff(
+                conversation_id,
                &event_turn_id,
                turn_diff_event,
                api_version,
@@ -574,12 +677,23 @@ pub(crate) async fn apply_bespoke_event_handling(
            )
            .await;
        }
+        EventMsg::PlanUpdate(plan_update_event) => {
+            handle_turn_plan_update(
+                conversation_id,
+                &event_turn_id,
+                plan_update_event,
+                api_version,
+                outgoing.as_ref(),
+            )
+            .await;
+        }

        _ => {}
    }
 }

 async fn handle_turn_diff(
+    conversation_id: ConversationId,
    event_turn_id: &str,
    turn_diff_event: TurnDiffEvent,
    api_version: ApiVersion,
@@ -587,6 +701,7 @@ async fn handle_turn_diff(
 ) {
    if let ApiVersion::V2 = api_version {
        let notification = TurnDiffUpdatedNotification {
+            thread_id: conversation_id.to_string(),
            turn_id: event_turn_id.to_string(),
            diff: turn_diff_event.unified_diff,
        };
@@ -596,10 +711,35 @@ async fn handle_turn_diff(
    }
 }

+async fn handle_turn_plan_update(
+    conversation_id: ConversationId,
+    event_turn_id: &str,
+    plan_update_event: UpdatePlanArgs,
+    api_version: ApiVersion,
+    outgoing: &OutgoingMessageSender,
+) {
+    if let ApiVersion::V2 = api_version {
+        let notification = TurnPlanUpdatedNotification {
+            thread_id: conversation_id.to_string(),
+            turn_id: event_turn_id.to_string(),
+            explanation: plan_update_event.explanation,
+            plan: plan_update_event
+                .plan
+                .into_iter()
+                .map(TurnPlanStep::from)
+                .collect(),
+        };
+        outgoing
+            .send_server_notification(ServerNotification::TurnPlanUpdated(notification))
+            .await;
+    }
+}
+
 async fn emit_turn_completed_with_status(
    conversation_id: ConversationId,
    event_turn_id: String,
    status: TurnStatus,
+    error: Option<TurnError>,
    outgoing: &OutgoingMessageSender,
 ) {
    let notification = TurnCompletedNotification {
@@ -607,6 +747,7 @@ async fn emit_turn_completed_with_status(
        turn: Turn {
            id: event_turn_id,
            items: vec![],
+            error,
            status,
        },
    };
@@ -695,13 +836,12 @@ async fn handle_turn_complete(
 ) {
    let turn_summary = find_and_remove_turn_summary(conversation_id, turn_summary_store).await;

-    let status = if let Some(error) = turn_summary.last_error {
-        TurnStatus::Failed { error }
-    } else {
-        TurnStatus::Completed
+    let (status, error) = match turn_summary.last_error {
+        Some(error) => (TurnStatus::Failed, Some(error)),
+        None => (TurnStatus::Completed, None),
    };

-    emit_turn_completed_with_status(conversation_id, event_turn_id, status, outgoing).await;
+    emit_turn_completed_with_status(conversation_id, event_turn_id, status, error, outgoing).await;
 }

 async fn handle_turn_interrupted(
@@ -716,6 +856,7 @@ async fn handle_turn_interrupted(
        conversation_id,
        event_turn_id,
        TurnStatus::Interrupted,
+        None,
        outgoing,
    )
    .await;
@@ -921,7 +1062,11 @@ async fn on_file_change_request_approval_response(
                });

            let (decision, completion_status) = match response.decision {
-                ApprovalDecision::Accept => (ReviewDecision::Approved, None),
+                ApprovalDecision::Accept
+                | ApprovalDecision::AcceptForSession
+                | ApprovalDecision::AcceptWithExecpolicyAmendment { .. } => {
+                    (ReviewDecision::Approved, None)
+                }
                ApprovalDecision::Decline => {
                    (ReviewDecision::Denied, Some(PatchApplyStatus::Declined))
                }
@@ -983,25 +1128,27 @@ async fn on_command_execution_request_approval_response(
                    error!("failed to deserialize CommandExecutionRequestApprovalResponse: {err}");
                    CommandExecutionRequestApprovalResponse {
                        decision: ApprovalDecision::Decline,
-                        accept_settings: None,
                    }
                });

-            let CommandExecutionRequestApprovalResponse {
-                decision,
-                accept_settings,
-            } = response;
+            let decision = response.decision;

-            let (decision, completion_status) = match (decision, accept_settings) {
-                (ApprovalDecision::Accept, Some(settings)) if settings.for_session => {
-                    (ReviewDecision::ApprovedForSession, None)
-                }
-                (ApprovalDecision::Accept, _) => (ReviewDecision::Approved, None),
-                (ApprovalDecision::Decline, _) => (
+            let (decision, completion_status) = match decision {
+                ApprovalDecision::Accept => (ReviewDecision::Approved, None),
+                ApprovalDecision::AcceptForSession => (ReviewDecision::ApprovedForSession, None),
+                ApprovalDecision::AcceptWithExecpolicyAmendment {
+                    execpolicy_amendment,
+                } => (
+                    ReviewDecision::ApprovedExecpolicyAmendment {
+                        proposed_execpolicy_amendment: execpolicy_amendment.into_core(),
+                    },
+                    None,
+                ),
+                ApprovalDecision::Decline => (
                    ReviewDecision::Denied,
                    Some(CommandExecutionStatus::Declined),
                ),
-                (ApprovalDecision::Cancel, _) => (
+                ApprovalDecision::Cancel => (
                    ReviewDecision::Abort,
                    Some(CommandExecutionStatus::Declined),
                ),
@@ -1054,6 +1201,7 @@ async fn construct_mcp_tool_call_notification(
        arguments: begin_event.invocation.arguments.unwrap_or(JsonValue::Null),
        result: None,
        error: None,
+        duration_ms: None,
    };
    ItemStartedNotification {
        thread_id,
@@ -1062,7 +1210,7 @@ async fn construct_mcp_tool_call_notification(
    }
 }

-/// simiilar to handle_mcp_tool_call_end in exec
+/// similar to handle_mcp_tool_call_end in exec
 async fn construct_mcp_tool_call_end_notification(
    end_event: McpToolCallEndEvent,
    thread_id: String,
@@ -1073,6 +1221,7 @@ async fn construct_mcp_tool_call_end_notification(
    } else {
        McpToolCallStatus::Failed
    };
+    let duration_ms = i64::try_from(end_event.duration.as_millis()).ok();

    let (result, error) = match &end_event.result {
        Ok(value) => (
@@ -1098,6 +1247,7 @@ async fn construct_mcp_tool_call_end_notification(
        arguments: end_event.invocation.arguments.unwrap_or(JsonValue::Null),
        result,
        error,
+        duration_ms,
    };
    ItemCompletedNotification {
        thread_id,
@@ -1115,12 +1265,15 @@ mod tests {
    use anyhow::Result;
    use anyhow::anyhow;
    use anyhow::bail;
+    use codex_app_server_protocol::TurnPlanStepStatus;
    use codex_core::protocol::CreditsSnapshot;
    use codex_core::protocol::McpInvocation;
    use codex_core::protocol::RateLimitSnapshot;
    use codex_core::protocol::RateLimitWindow;
    use codex_core::protocol::TokenUsage;
    use codex_core::protocol::TokenUsageInfo;
+    use codex_protocol::plan_tool::PlanItemArg;
+    use codex_protocol::plan_tool::StepStatus;
    use mcp_types::CallToolResult;
    use mcp_types::ContentBlock;
    use mcp_types::TextContent;
@@ -1185,6 +1338,7 @@ mod tests {
            OutgoingMessage::AppServerNotification(ServerNotification::TurnCompleted(n)) => {
                assert_eq!(n.turn.id, event_turn_id);
                assert_eq!(n.turn.status, TurnStatus::Completed);
+                assert_eq!(n.turn.error, None);
            }
            other => bail!("unexpected message: {other:?}"),
        }
@@ -1225,6 +1379,7 @@ mod tests {
            OutgoingMessage::AppServerNotification(ServerNotification::TurnCompleted(n)) => {
                assert_eq!(n.turn.id, event_turn_id);
                assert_eq!(n.turn.status, TurnStatus::Interrupted);
+                assert_eq!(n.turn.error, None);
            }
            other => bail!("unexpected message: {other:?}"),
        }
@@ -1264,14 +1419,13 @@ mod tests {
        match msg {
            OutgoingMessage::AppServerNotification(ServerNotification::TurnCompleted(n)) => {
                assert_eq!(n.turn.id, event_turn_id);
+                assert_eq!(n.turn.status, TurnStatus::Failed);
                assert_eq!(
-                    n.turn.status,
-                    TurnStatus::Failed {
-                        error: TurnError {
-                            message: "bad".to_string(),
-                            codex_error_info: Some(V2CodexErrorInfo::Other),
-                        }
-                    }
+                    n.turn.error,
+                    Some(TurnError {
+                        message: "bad".to_string(),
+                        codex_error_info: Some(V2CodexErrorInfo::Other),
+                    })
                );
            }
            other => bail!("unexpected message: {other:?}"),
@@ -1280,6 +1434,56 @@ mod tests {
        Ok(())
    }

+    #[tokio::test]
+    async fn test_handle_turn_plan_update_emits_notification_for_v2() -> Result<()> {
+        let (tx, mut rx) = mpsc::channel(CHANNEL_CAPACITY);
+        let outgoing = OutgoingMessageSender::new(tx);
+        let update = UpdatePlanArgs {
+            explanation: Some("need plan".to_string()),
+            plan: vec![
+                PlanItemArg {
+                    step: "first".to_string(),
+                    status: StepStatus::Pending,
+                },
+                PlanItemArg {
+                    step: "second".to_string(),
+                    status: StepStatus::Completed,
+                },
+            ],
+        };
+
+        let conversation_id = ConversationId::new();
+
+        handle_turn_plan_update(
+            conversation_id,
+            "turn-123",
+            update,
+            ApiVersion::V2,
+            &outgoing,
+        )
+        .await;
+
+        let msg = rx
+            .recv()
+            .await
+            .ok_or_else(|| anyhow!("should send one notification"))?;
+        match msg {
+            OutgoingMessage::AppServerNotification(ServerNotification::TurnPlanUpdated(n)) => {
+                assert_eq!(n.thread_id, conversation_id.to_string());
+                assert_eq!(n.turn_id, "turn-123");
+                assert_eq!(n.explanation.as_deref(), Some("need plan"));
+                assert_eq!(n.plan.len(), 2);
+                assert_eq!(n.plan[0].step, "first");
+                assert_eq!(n.plan[0].status, TurnPlanStepStatus::Pending);
+                assert_eq!(n.plan[1].step, "second");
+                assert_eq!(n.plan[1].status, TurnPlanStepStatus::Completed);
+            }
+            other => bail!("unexpected message: {other:?}"),
+        }
+        assert!(rx.try_recv().is_err(), "no extra messages expected");
+        Ok(())
+    }
+
    #[tokio::test]
    async fn test_handle_token_count_event_emits_usage_and_rate_limits() -> Result<()> {
        let conversation_id = ConversationId::new();
@@ -1316,6 +1520,7 @@ mod tests {
                unlimited: false,
                balance: Some("5".to_string()),
            }),
+            plan_type: None,
        };

        handle_token_count_event(
@@ -1420,6 +1625,7 @@ mod tests {
                arguments: serde_json::json!({"server": ""}),
                result: None,
                error: None,
+                duration_ms: None,
            },
        };

@@ -1492,14 +1698,13 @@ mod tests {
        match msg {
            OutgoingMessage::AppServerNotification(ServerNotification::TurnCompleted(n)) => {
                assert_eq!(n.turn.id, a_turn1);
+                assert_eq!(n.turn.status, TurnStatus::Failed);
                assert_eq!(
-                    n.turn.status,
-                    TurnStatus::Failed {
-                        error: TurnError {
-                            message: "a1".to_string(),
-                            codex_error_info: Some(V2CodexErrorInfo::BadRequest),
-                        }
-                    }
+                    n.turn.error,
+                    Some(TurnError {
+                        message: "a1".to_string(),
+                        codex_error_info: Some(V2CodexErrorInfo::BadRequest),
+                    })
                );
            }
            other => bail!("unexpected message: {other:?}"),
@@ -1513,14 +1718,13 @@ mod tests {
        match msg {
            OutgoingMessage::AppServerNotification(ServerNotification::TurnCompleted(n)) => {
                assert_eq!(n.turn.id, b_turn1);
+                assert_eq!(n.turn.status, TurnStatus::Failed);
                assert_eq!(
-                    n.turn.status,
-                    TurnStatus::Failed {
-                        error: TurnError {
-                            message: "b1".to_string(),
-                            codex_error_info: None,
-                        }
-                    }
+                    n.turn.error,
+                    Some(TurnError {
+                        message: "b1".to_string(),
+                        codex_error_info: None,
+                    })
                );
            }
            other => bail!("unexpected message: {other:?}"),
@@ -1535,6 +1739,7 @@ mod tests {
            OutgoingMessage::AppServerNotification(ServerNotification::TurnCompleted(n)) => {
                assert_eq!(n.turn.id, a_turn2);
                assert_eq!(n.turn.status, TurnStatus::Completed);
+                assert_eq!(n.turn.error, None);
            }
            other => bail!("unexpected message: {other:?}"),
        }
@@ -1574,6 +1779,7 @@ mod tests {
                arguments: JsonValue::Null,
                result: None,
                error: None,
+                duration_ms: None,
            },
        };

@@ -1627,6 +1833,7 @@ mod tests {
                    structured_content: None,
                }),
                error: None,
+                duration_ms: Some(0),
            },
        };

@@ -1668,6 +1875,7 @@ mod tests {
                error: Some(McpToolCallError {
                    message: "boom".to_string(),
                }),
+                duration_ms: Some(1),
            },
        };

@@ -1679,8 +1887,10 @@ mod tests {
        let (tx, mut rx) = mpsc::channel(CHANNEL_CAPACITY);
        let outgoing = OutgoingMessageSender::new(tx);
        let unified_diff = "--- a\n+++ b\n".to_string();
+        let conversation_id = ConversationId::new();

        handle_turn_diff(
+            conversation_id,
            "turn-1",
            TurnDiffEvent {
                unified_diff: unified_diff.clone(),
@@ -1698,6 +1908,7 @@ mod tests {
            OutgoingMessage::AppServerNotification(ServerNotification::TurnDiffUpdated(
                notification,
            )) => {
+                assert_eq!(notification.thread_id, conversation_id.to_string());
                assert_eq!(notification.turn_id, "turn-1");
                assert_eq!(notification.diff, unified_diff);
            }
@@ -1711,8 +1922,10 @@ mod tests {
    async fn test_handle_turn_diff_is_noop_for_v1() -> Result<()> {
        let (tx, mut rx) = mpsc::channel(CHANNEL_CAPACITY);
        let outgoing = OutgoingMessageSender::new(tx);
+        let conversation_id = ConversationId::new();

        handle_turn_diff(
+            conversation_id,
            "turn-1",
            TurnDiffEvent {
                unified_diff: "diff".to_string(),
--- a/codex-rs/app-server/src/codex_message_processor.rs
+++ b/codex-rs/app-server/src/codex_message_processor.rs
--- a/codex-rs/app-server/src/config_api.rs
+++ b/codex-rs/app-server/src/config_api.rs
@@ -1,6 +1,6 @@
 use crate::error_code::INTERNAL_ERROR_CODE;
 use crate::error_code::INVALID_REQUEST_ERROR_CODE;
-use anyhow::anyhow;
+use codex_app_server_protocol::Config;
 use codex_app_server_protocol::ConfigBatchWriteParams;
 use codex_app_server_protocol::ConfigLayer;
 use codex_app_server_protocol::ConfigLayerMetadata;
@@ -15,6 +15,8 @@ use codex_app_server_protocol::MergeStrategy;
 use codex_app_server_protocol::OverriddenMetadata;
 use codex_app_server_protocol::WriteStatus;
 use codex_core::config::ConfigToml;
+use codex_core::config::edit::ConfigEdit;
+use codex_core::config::edit::ConfigEditsBuilder;
 use codex_core::config_loader::LoadedConfigLayers;
 use codex_core::config_loader::LoaderOverrides;
 use codex_core::config_loader::load_config_layers_with_overrides;
@@ -26,9 +28,8 @@ use sha2::Sha256;
 use std::collections::HashMap;
 use std::path::Path;
 use std::path::PathBuf;
-use tempfile::NamedTempFile;
-use tokio::task;
 use toml::Value as TomlValue;
+use toml_edit::Item as TomlItem;

 const SESSION_FLAGS_SOURCE: &str = "--config";
 const MDM_SOURCE: &str = "com.openai.codex/config_toml_base64";
@@ -75,8 +76,10 @@ impl ConfigApi {
        let effective = layers.effective_config();
        validate_config(&effective).map_err(|err| internal_error("invalid configuration", err))?;

+        let config: Config = serde_json::from_value(to_json_value(&effective))
+            .map_err(|err| internal_error("failed to deserialize configuration", err))?;
        let response = ConfigReadResponse {
-            config: to_json_value(&effective),
+            config,
            origins: layers.origins(),
            layers: params.include_layers.then(|| layers.layers_high_to_low()),
        };
@@ -109,12 +112,17 @@ impl ConfigApi {

    async fn apply_edits(
        &self,
-        file_path: String,
+        file_path: Option<String>,
        expected_version: Option<String>,
        edits: Vec<(String, JsonValue, MergeStrategy)>,
    ) -> Result<ConfigWriteResponse, JSONRPCErrorError> {
        let allowed_path = self.codex_home.join(CONFIG_FILE_NAME);
-        if !paths_match(&allowed_path, &file_path) {
+        let provided_path = file_path
+            .as_ref()
+            .map(PathBuf::from)
+            .unwrap_or_else(|| allowed_path.clone());
+
+        if !paths_match(&allowed_path, &provided_path) {
            return Err(config_write_error(
                ConfigWriteErrorCode::ConfigLayerReadonly,
                "Only writes to the user config are allowed",
@@ -136,19 +144,20 @@ impl ConfigApi {
        }

        let mut user_config = layers.user.config.clone();
-        let mut mutated = false;
        let mut parsed_segments = Vec::new();
+        let mut config_edits = Vec::new();

        for (key_path, value, strategy) in edits.into_iter() {
            let segments = parse_key_path(&key_path).map_err(|message| {
                config_write_error(ConfigWriteErrorCode::ConfigValidationError, message)
            })?;
+            let original_value = value_at_path(&user_config, &segments).cloned();
            let parsed_value = parse_value(value).map_err(|message| {
                config_write_error(ConfigWriteErrorCode::ConfigValidationError, message)
            })?;

-            let changed = apply_merge(&mut user_config, &segments, parsed_value.as_ref(), strategy)
-                .map_err(|err| match err {
+            apply_merge(&mut user_config, &segments, parsed_value.as_ref(), strategy).map_err(
+                |err| match err {
                    MergeError::PathNotFound => config_write_error(
                        ConfigWriteErrorCode::ConfigPathNotFound,
                        "Path not found",
@@ -156,9 +165,24 @@ impl ConfigApi {
                    MergeError::Validation(message) => {
                        config_write_error(ConfigWriteErrorCode::ConfigValidationError, message)
                    }
-                })?;
+                },
+            )?;
+
+            let updated_value = value_at_path(&user_config, &segments).cloned();
+            if original_value != updated_value {
+                let edit = match updated_value {
+                    Some(value) => ConfigEdit::SetPath {
+                        segments: segments.clone(),
+                        value: toml_value_to_item(&value)
+                            .map_err(|err| internal_error("failed to build config edits", err))?,
+                    },
+                    None => ConfigEdit::ClearPath {
+                        segments: segments.clone(),
+                    },
+                };
+                config_edits.push(edit);
+            }

-            mutated |= changed;
            parsed_segments.push(segments);
        }

@@ -178,8 +202,10 @@ impl ConfigApi {
            )
        })?;

-        if mutated {
-            self.persist_user_config(&user_config)
+        if !config_edits.is_empty() {
+            ConfigEditsBuilder::new(&self.codex_home)
+                .with_edits(config_edits)
+                .apply()
                .await
                .map_err(|err| internal_error("failed to persist config.toml", err))?;
        }
@@ -190,9 +216,16 @@ impl ConfigApi {
            .map(|_| WriteStatus::OkOverridden)
            .unwrap_or(WriteStatus::Ok);

+        let file_path = provided_path
+            .canonicalize()
+            .unwrap_or(provided_path.clone())
+            .display()
+            .to_string();
+
        Ok(ConfigWriteResponse {
            status,
            version: updated_layers.user.version.clone(),
+            file_path,
            overridden_metadata: overridden,
        })
    }
@@ -241,25 +274,6 @@ impl ConfigApi {
            mdm,
        })
    }
-
-    async fn persist_user_config(&self, user_config: &TomlValue) -> anyhow::Result<()> {
-        let codex_home = self.codex_home.clone();
-        let serialized = toml::to_string_pretty(user_config)?;
-
-        task::spawn_blocking(move || -> anyhow::Result<()> {
-            std::fs::create_dir_all(&codex_home)?;
-
-            let target = codex_home.join(CONFIG_FILE_NAME);
-            let tmp = NamedTempFile::new_in(&codex_home)?;
-            std::fs::write(tmp.path(), serialized.as_bytes())?;
-            tmp.persist(&target)?;
-            Ok(())
-        })
-        .await
-        .map_err(|err| anyhow!("config persistence task panicked: {err}"))??;
-
-        Ok(())
-    }
 }

 fn parse_value(value: JsonValue) -> Result<Option<TomlValue>, String> {
@@ -410,6 +424,44 @@ fn clear_path(root: &mut TomlValue, segments: &[String]) -> Result<bool, MergeEr
    Ok(parent.remove(last).is_some())
 }

+fn toml_value_to_item(value: &TomlValue) -> anyhow::Result<TomlItem> {
+    match value {
+        TomlValue::Table(table) => {
+            let mut table_item = toml_edit::Table::new();
+            table_item.set_implicit(false);
+            for (key, val) in table {
+                table_item.insert(key, toml_value_to_item(val)?);
+            }
+            Ok(TomlItem::Table(table_item))
+        }
+        other => Ok(TomlItem::Value(toml_value_to_value(other)?)),
+    }
+}
+
+fn toml_value_to_value(value: &TomlValue) -> anyhow::Result<toml_edit::Value> {
+    match value {
+        TomlValue::String(val) => Ok(toml_edit::Value::from(val.clone())),
+        TomlValue::Integer(val) => Ok(toml_edit::Value::from(*val)),
+        TomlValue::Float(val) => Ok(toml_edit::Value::from(*val)),
+        TomlValue::Boolean(val) => Ok(toml_edit::Value::from(*val)),
+        TomlValue::Datetime(val) => Ok(toml_edit::Value::from(*val)),
+        TomlValue::Array(items) => {
+            let mut array = toml_edit::Array::new();
+            for item in items {
+                array.push(toml_value_to_value(item)?);
+            }
+            Ok(toml_edit::Value::Array(array))
+        }
+        TomlValue::Table(table) => {
+            let mut inline = toml_edit::InlineTable::new();
+            for (key, val) in table {
+                inline.insert(key, toml_value_to_value(val)?);
+            }
+            Ok(toml_edit::Value::InlineTable(inline))
+        }
+    }
+}
+
 #[derive(Clone)]
 struct LayerState {
    name: ConfigLayerName,
@@ -587,15 +639,14 @@ fn canonical_json(value: &JsonValue) -> JsonValue {
    }
 }

-fn paths_match(expected: &Path, provided: &str) -> bool {
-    let provided_path = PathBuf::from(provided);
+fn paths_match(expected: &Path, provided: &Path) -> bool {
    if let (Ok(expanded_expected), Ok(expanded_provided)) =
-        (expected.canonicalize(), provided_path.canonicalize())
+        (expected.canonicalize(), provided.canonicalize())
    {
        return expanded_expected == expanded_provided;
    }

-    expected == provided_path
+    expected == provided
 }

 fn value_at_path<'a>(root: &'a TomlValue, segments: &[String]) -> Option<&'a TomlValue> {
@@ -724,9 +775,105 @@ fn config_write_error(code: ConfigWriteErrorCode, message: impl Into<String>) ->
 #[cfg(test)]
 mod tests {
    use super::*;
+    use anyhow::Result;
+    use codex_app_server_protocol::AskForApproval;
    use pretty_assertions::assert_eq;
    use tempfile::tempdir;

+    #[test]
+    fn toml_value_to_item_handles_nested_config_tables() {
+        let config = r#"
+[mcp_servers.docs]
+command = "docs-server"
+
+[mcp_servers.docs.http_headers]
+X-Doc = "42"
+"#;
+
+        let value: TomlValue = toml::from_str(config).expect("parse config example");
+        let item = toml_value_to_item(&value).expect("convert to toml_edit item");
+
+        let root = item.as_table().expect("root table");
+        assert!(!root.is_implicit(), "root table should be explicit");
+
+        let mcp_servers = root
+            .get("mcp_servers")
+            .and_then(TomlItem::as_table)
+            .expect("mcp_servers table");
+        assert!(
+            !mcp_servers.is_implicit(),
+            "mcp_servers table should be explicit"
+        );
+
+        let docs = mcp_servers
+            .get("docs")
+            .and_then(TomlItem::as_table)
+            .expect("docs table");
+        assert_eq!(
+            docs.get("command")
+                .and_then(TomlItem::as_value)
+                .and_then(toml_edit::Value::as_str),
+            Some("docs-server")
+        );
+
+        let http_headers = docs
+            .get("http_headers")
+            .and_then(TomlItem::as_table)
+            .expect("http_headers table");
+        assert_eq!(
+            http_headers
+                .get("X-Doc")
+                .and_then(TomlItem::as_value)
+                .and_then(toml_edit::Value::as_str),
+            Some("42")
+        );
+    }
+
+    #[tokio::test]
+    async fn write_value_preserves_comments_and_order() -> Result<()> {
+        let tmp = tempdir().expect("tempdir");
+        let original = r#"# Codex user configuration
+model = "gpt-5"
+approval_policy = "on-request"
+
+[notice]
+# Preserve this comment
+hide_full_access_warning = true
+
+[features]
+unified_exec = true
+"#;
+        std::fs::write(tmp.path().join(CONFIG_FILE_NAME), original)?;
+
+        let api = ConfigApi::new(tmp.path().to_path_buf(), vec![]);
+        api.write_value(ConfigValueWriteParams {
+            file_path: Some(tmp.path().join(CONFIG_FILE_NAME).display().to_string()),
+            key_path: "features.remote_compaction".to_string(),
+            value: json!(true),
+            merge_strategy: MergeStrategy::Replace,
+            expected_version: None,
+        })
+        .await
+        .expect("write succeeds");
+
+        let updated =
+            std::fs::read_to_string(tmp.path().join(CONFIG_FILE_NAME)).expect("read config");
+        let expected = r#"# Codex user configuration
+model = "gpt-5"
+approval_policy = "on-request"
+
+[notice]
+# Preserve this comment
+hide_full_access_warning = true
+
+[features]
+unified_exec = true
+remote_compaction = true
+"#;
+        assert_eq!(updated, expected);
+        Ok(())
+    }
+
    #[tokio::test]
    async fn read_includes_origins_and_layers() {
        let tmp = tempdir().expect("tempdir");
@@ -752,10 +899,7 @@ mod tests {
            .await
            .expect("response");

-        assert_eq!(
-            response.config.get("approval_policy"),
-            Some(&json!("never"))
-        );
+        assert_eq!(response.config.approval_policy, Some(AskForApproval::Never));

        assert_eq!(
            response
@@ -795,7 +939,7 @@ mod tests {

        let result = api
            .write_value(ConfigValueWriteParams {
-                file_path: tmp.path().join(CONFIG_FILE_NAME).display().to_string(),
+                file_path: Some(tmp.path().join(CONFIG_FILE_NAME).display().to_string()),
                key_path: "approval_policy".to_string(),
                value: json!("never"),
                merge_strategy: MergeStrategy::Replace,
@@ -810,8 +954,10 @@ mod tests {
            })
            .await
            .expect("read");
-        let config_object = read_after.config.as_object().expect("object");
-        assert_eq!(config_object.get("approval_policy"), Some(&json!("never")));
+        assert_eq!(
+            read_after.config.approval_policy,
+            Some(AskForApproval::Never)
+        );
        assert_eq!(
            read_after
                .origins
@@ -832,7 +978,7 @@ mod tests {
        let api = ConfigApi::new(tmp.path().to_path_buf(), vec![]);
        let error = api
            .write_value(ConfigValueWriteParams {
-                file_path: tmp.path().join(CONFIG_FILE_NAME).display().to_string(),
+                file_path: Some(tmp.path().join(CONFIG_FILE_NAME).display().to_string()),
                key_path: "model".to_string(),
                value: json!("gpt-5"),
                merge_strategy: MergeStrategy::Replace,
@@ -852,6 +998,30 @@ mod tests {
        );
    }

+    #[tokio::test]
+    async fn write_value_defaults_to_user_config_path() {
+        let tmp = tempdir().expect("tempdir");
+        std::fs::write(tmp.path().join(CONFIG_FILE_NAME), "").unwrap();
+
+        let api = ConfigApi::new(tmp.path().to_path_buf(), vec![]);
+        api.write_value(ConfigValueWriteParams {
+            file_path: None,
+            key_path: "model".to_string(),
+            value: json!("gpt-new"),
+            merge_strategy: MergeStrategy::Replace,
+            expected_version: None,
+        })
+        .await
+        .expect("write succeeds");
+
+        let contents =
+            std::fs::read_to_string(tmp.path().join(CONFIG_FILE_NAME)).expect("read config");
+        assert!(
+            contents.contains("model = \"gpt-new\""),
+            "config.toml should be updated even when file_path is omitted"
+        );
+    }
+
    #[tokio::test]
    async fn invalid_user_value_rejected_even_if_overridden_by_managed() {
        let tmp = tempdir().expect("tempdir");
@@ -872,7 +1042,7 @@ mod tests {

        let error = api
            .write_value(ConfigValueWriteParams {
-                file_path: tmp.path().join(CONFIG_FILE_NAME).display().to_string(),
+                file_path: Some(tmp.path().join(CONFIG_FILE_NAME).display().to_string()),
                key_path: "approval_policy".to_string(),
                value: json!("bogus"),
                merge_strategy: MergeStrategy::Replace,
@@ -926,7 +1096,7 @@ mod tests {
            .await
            .expect("response");

-        assert_eq!(response.config.get("model"), Some(&json!("system")));
+        assert_eq!(response.config.model.as_deref(), Some("system"));
        assert_eq!(
            response.origins.get("model").expect("origin").name,
            ConfigLayerName::System
@@ -957,7 +1127,7 @@ mod tests {

        let result = api
            .write_value(ConfigValueWriteParams {
-                file_path: tmp.path().join(CONFIG_FILE_NAME).display().to_string(),
+                file_path: Some(tmp.path().join(CONFIG_FILE_NAME).display().to_string()),
                key_path: "approval_policy".to_string(),
                value: json!("on-request"),
                merge_strategy: MergeStrategy::Replace,
--- a/codex-rs/app-server/src/message_processor.rs
+++ b/codex-rs/app-server/src/message_processor.rs
@@ -59,6 +59,7 @@ impl MessageProcessor {
            outgoing.clone(),
            codex_linux_sandbox_exe,
            Arc::clone(&config),
+            cli_overrides.clone(),
            feedback,
        );
        let config_api = ConfigApi::new(config.codex_home.clone(), cli_overrides);
--- a/codex-rs/app-server/src/models.rs
+++ b/codex-rs/app-server/src/models.rs
@@ -1,12 +1,19 @@
-use codex_app_server_protocol::AuthMode;
+use std::sync::Arc;
+
 use codex_app_server_protocol::Model;
 use codex_app_server_protocol::ReasoningEffortOption;
-use codex_common::model_presets::ModelPreset;
-use codex_common::model_presets::ReasoningEffortPreset;
-use codex_common::model_presets::builtin_model_presets;
+use codex_core::ConversationManager;
+use codex_core::config::Config;
+use codex_protocol::openai_models::ModelPreset;
+use codex_protocol::openai_models::ReasoningEffortPreset;

-pub fn supported_models(auth_mode: Option<AuthMode>) -> Vec<Model> {
-    builtin_model_presets(auth_mode)
+pub async fn supported_models(
+    conversation_manager: Arc<ConversationManager>,
+    config: &Config,
+) -> Vec<Model> {
+    conversation_manager
+        .list_models(config)
+        .await
        .into_iter()
        .map(model_from_preset)
        .collect()
@@ -27,7 +34,7 @@ fn model_from_preset(preset: ModelPreset) -> Model {
 }

 fn reasoning_efforts_from_preset(
-    efforts: &'static [ReasoningEffortPreset],
+    efforts: Vec<ReasoningEffortPreset>,
 ) -> Vec<ReasoningEffortOption> {
    efforts
        .iter()
--- a/codex-rs/app-server/src/outgoing_message.rs
+++ b/codex-rs/app-server/src/outgoing_message.rs
@@ -16,6 +16,9 @@ use tracing::warn;

 use crate::error_code::INTERNAL_ERROR_CODE;

+#[cfg(test)]
+use codex_protocol::account::PlanType;
+
 /// Sends messages to the client and manages request callbacks.
 pub(crate) struct OutgoingMessageSender {
    next_request_id: AtomicI64,
@@ -230,6 +233,7 @@ mod tests {
                    }),
                    secondary: None,
                    credits: None,
+                    plan_type: Some(PlanType::Plus),
                },
            });

@@ -245,7 +249,8 @@ mod tests {
                            "resetsAt": 123
                        },
                        "secondary": null,
-                        "credits": null
+                        "credits": null,
+                        "planType": "plus"
                    }
                },
            }),
--- a/codex-rs/app-server/tests/common/lib.rs
+++ b/codex-rs/app-server/tests/common/lib.rs
@@ -1,6 +1,7 @@
 mod auth_fixtures;
 mod mcp_process;
 mod mock_model_server;
+mod models_cache;
 mod responses;
 mod rollout;

@@ -11,9 +12,13 @@ pub use auth_fixtures::write_chatgpt_auth;
 use codex_app_server_protocol::JSONRPCResponse;
 pub use core_test_support::format_with_current_shell;
 pub use core_test_support::format_with_current_shell_display;
+pub use core_test_support::format_with_current_shell_display_non_login;
+pub use core_test_support::format_with_current_shell_non_login;
 pub use mcp_process::McpProcess;
 pub use mock_model_server::create_mock_chat_completions_server;
 pub use mock_model_server::create_mock_chat_completions_server_unchecked;
+pub use models_cache::write_models_cache;
+pub use models_cache::write_models_cache_with_models;
 pub use responses::create_apply_patch_sse_response;
 pub use responses::create_exec_command_sse_response;
 pub use responses::create_final_assistant_message_sse_response;
--- a/codex-rs/app-server/tests/common/models_cache.rs
+++ b/codex-rs/app-server/tests/common/models_cache.rs
@@ -0,0 +1,74 @@
+use chrono::DateTime;
+use chrono::Utc;
+use codex_core::openai_models::model_presets::all_model_presets;
+use codex_protocol::openai_models::ClientVersion;
+use codex_protocol::openai_models::ConfigShellToolType;
+use codex_protocol::openai_models::ModelInfo;
+use codex_protocol::openai_models::ModelPreset;
+use codex_protocol::openai_models::ModelVisibility;
+use serde_json::json;
+use std::path::Path;
+
+/// Convert a ModelPreset to ModelInfo for cache storage.
+fn preset_to_info(preset: &ModelPreset, priority: i32) -> ModelInfo {
+    ModelInfo {
+        slug: preset.id.clone(),
+        display_name: preset.display_name.clone(),
+        description: Some(preset.description.clone()),
+        default_reasoning_level: preset.default_reasoning_effort,
+        supported_reasoning_levels: preset.supported_reasoning_efforts.clone(),
+        shell_type: ConfigShellToolType::ShellCommand,
+        visibility: if preset.show_in_picker {
+            ModelVisibility::List
+        } else {
+            ModelVisibility::Hide
+        },
+        minimal_client_version: ClientVersion(0, 1, 0),
+        supported_in_api: true,
+        priority,
+        upgrade: preset.upgrade.as_ref().map(|u| u.id.clone()),
+        base_instructions: None,
+    }
+}
+
+/// Write a models_cache.json file to the codex home directory.
+/// This prevents ModelsManager from making network requests to refresh models.
+/// The cache will be treated as fresh (within TTL) and used instead of fetching from the network.
+/// Uses the built-in model presets from ModelsManager, converted to ModelInfo format.
+pub fn write_models_cache(codex_home: &Path) -> std::io::Result<()> {
+    // Get all presets and filter for show_in_picker (same as builtin_model_presets does)
+    let presets: Vec<&ModelPreset> = all_model_presets()
+        .iter()
+        .filter(|preset| preset.show_in_picker)
+        .collect();
+    // Convert presets to ModelInfo, assigning priorities (higher = earlier in list)
+    // Priority is used for sorting, so first model gets highest priority
+    let models: Vec<ModelInfo> = presets
+        .iter()
+        .enumerate()
+        .map(|(idx, preset)| {
+            // Higher priority = earlier in list, so reverse the index
+            let priority = (presets.len() - idx) as i32;
+            preset_to_info(preset, priority)
+        })
+        .collect();
+
+    write_models_cache_with_models(codex_home, models)
+}
+
+/// Write a models_cache.json file with specific models.
+/// Useful when tests need specific models to be available.
+pub fn write_models_cache_with_models(
+    codex_home: &Path,
+    models: Vec<ModelInfo>,
+) -> std::io::Result<()> {
+    let cache_path = codex_home.join("models_cache.json");
+    // DateTime<Utc> serializes to RFC3339 format by default with serde
+    let fetched_at: DateTime<Utc> = Utc::now();
+    let cache = json!({
+        "fetched_at": fetched_at,
+        "etag": null,
+        "models": models
+    });
+    std::fs::write(cache_path, serde_json::to_string_pretty(&cache)?)
+}
--- a/codex-rs/app-server/tests/suite/codex_message_processor_flow.rs
+++ b/codex-rs/app-server/tests/suite/codex_message_processor_flow.rs
@@ -23,10 +23,10 @@ use codex_app_server_protocol::SendUserTurnResponse;
 use codex_app_server_protocol::ServerRequest;
 use codex_core::protocol::AskForApproval;
 use codex_core::protocol::SandboxPolicy;
-use codex_core::protocol_config_types::ReasoningEffort;
 use codex_core::protocol_config_types::ReasoningSummary;
 use codex_core::spawn::CODEX_SANDBOX_NETWORK_DISABLED_ENV_VAR;
 use codex_protocol::config_types::SandboxMode;
+use codex_protocol::openai_models::ReasoningEffort;
 use codex_protocol::parse_command::ParsedCommand;
 use codex_protocol::protocol::Event;
 use codex_protocol::protocol::EventMsg;
@@ -271,7 +271,6 @@ async fn test_send_user_turn_changes_approval_policy_behavior() -> Result<()> {
            command: format_with_current_shell("python3 -c 'print(42)'"),
            cwd: working_directory.clone(),
            reason: None,
-            risk: None,
            parsed_cmd: vec![ParsedCommand::Unknown {
                cmd: "python3 -c 'print(42)'".to_string()
            }],
--- a/codex-rs/app-server/tests/suite/config.rs
+++ b/codex-rs/app-server/tests/suite/config.rs
@@ -10,10 +10,10 @@ use codex_app_server_protocol::Tools;
 use codex_app_server_protocol::UserSavedConfig;
 use codex_core::protocol::AskForApproval;
 use codex_protocol::config_types::ForcedLoginMethod;
-use codex_protocol::config_types::ReasoningEffort;
 use codex_protocol::config_types::ReasoningSummary;
 use codex_protocol::config_types::SandboxMode;
 use codex_protocol::config_types::Verbosity;
+use codex_protocol::openai_models::ReasoningEffort;
 use pretty_assertions::assert_eq;
 use std::collections::HashMap;
 use std::path::Path;
--- a/codex-rs/app-server/tests/suite/list_resume.rs
+++ b/codex-rs/app-server/tests/suite/list_resume.rs
@@ -358,3 +358,81 @@ async fn test_list_and_resume_conversations() -> Result<()> {

    Ok(())
 }
+
+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn list_conversations_fetches_through_filtered_pages() -> Result<()> {
+    let codex_home = TempDir::new()?;
+
+    // Only the last 3 conversations match the provider filter; request 3 and
+    // ensure pagination keeps fetching past non-matching pages.
+    let cases = [
+        (
+            "2025-03-04T12-00-00",
+            "2025-03-04T12:00:00Z",
+            "skip_provider",
+        ),
+        (
+            "2025-03-03T12-00-00",
+            "2025-03-03T12:00:00Z",
+            "skip_provider",
+        ),
+        (
+            "2025-03-02T12-00-00",
+            "2025-03-02T12:00:00Z",
+            "target_provider",
+        ),
+        (
+            "2025-03-01T12-00-00",
+            "2025-03-01T12:00:00Z",
+            "target_provider",
+        ),
+        (
+            "2025-02-28T12-00-00",
+            "2025-02-28T12:00:00Z",
+            "target_provider",
+        ),
+    ];
+
+    for (ts_file, ts_rfc, provider) in cases {
+        create_fake_rollout(
+            codex_home.path(),
+            ts_file,
+            ts_rfc,
+            "Hello",
+            Some(provider),
+            None,
+        )?;
+    }
+
+    let mut mcp = McpProcess::new(codex_home.path()).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+
+    let req_id = mcp
+        .send_list_conversations_request(ListConversationsParams {
+            page_size: Some(3),
+            cursor: None,
+            model_providers: Some(vec!["target_provider".to_string()]),
+        })
+        .await?;
+    let resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(req_id)),
+    )
+    .await??;
+    let ListConversationsResponse { items, next_cursor } =
+        to_response::<ListConversationsResponse>(resp)?;
+
+    assert_eq!(
+        items.len(),
+        3,
+        "should fetch across pages to satisfy the limit"
+    );
+    assert!(
+        items
+            .iter()
+            .all(|item| item.model_provider == "target_provider")
+    );
+    assert_eq!(next_cursor, None);
+
+    Ok(())
+}
--- a/codex-rs/app-server/tests/suite/login.rs
+++ b/codex-rs/app-server/tests/suite/login.rs
@@ -1,8 +1,6 @@
 use anyhow::Result;
 use app_test_support::McpProcess;
 use app_test_support::to_response;
-use codex_app_server_protocol::CancelLoginChatGptParams;
-use codex_app_server_protocol::CancelLoginChatGptResponse;
 use codex_app_server_protocol::GetAuthStatusParams;
 use codex_app_server_protocol::GetAuthStatusResponse;
 use codex_app_server_protocol::JSONRPCError;
@@ -14,7 +12,6 @@ use codex_core::auth::AuthCredentialsStoreMode;
 use codex_login::login_with_api_key;
 use serial_test::serial;
 use std::path::Path;
-use std::time::Duration;
 use tempfile::TempDir;
 use tokio::time::timeout;

@@ -87,48 +84,6 @@ async fn logout_chatgpt_removes_auth() -> Result<()> {
    Ok(())
 }

-#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
-// Serialize tests that launch the login server since it binds to a fixed port.
-#[serial(login_port)]
-async fn login_and_cancel_chatgpt() -> Result<()> {
-    let codex_home = TempDir::new()?;
-    create_config_toml(codex_home.path())?;
-
-    let mut mcp = McpProcess::new(codex_home.path()).await?;
-    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
-
-    let login_id = mcp.send_login_chat_gpt_request().await?;
-    let login_resp: JSONRPCResponse = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_response_message(RequestId::Integer(login_id)),
-    )
-    .await??;
-    let login: LoginChatGptResponse = to_response(login_resp)?;
-
-    let cancel_id = mcp
-        .send_cancel_login_chat_gpt_request(CancelLoginChatGptParams {
-            login_id: login.login_id,
-        })
-        .await?;
-    let cancel_resp: JSONRPCResponse = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_response_message(RequestId::Integer(cancel_id)),
-    )
-    .await??;
-    let _ok: CancelLoginChatGptResponse = to_response(cancel_resp)?;
-
-    // Optionally observe the completion notification; do not fail if it races.
-    let maybe_note = timeout(
-        Duration::from_secs(2),
-        mcp.read_stream_until_notification_message("codex/event/login_chat_gpt_complete"),
-    )
-    .await;
-    if maybe_note.is_err() {
-        eprintln!("warning: did not observe login_chat_gpt_complete notification after cancel");
-    }
-    Ok(())
-}
-
 fn create_config_toml_forced_login(codex_home: &Path, forced_method: &str) -> std::io::Result<()> {
    let config_toml = codex_home.join("config.toml");
    let contents = format!(
--- a/codex-rs/app-server/tests/suite/v2/account.rs
+++ b/codex-rs/app-server/tests/suite/v2/account.rs
@@ -241,7 +241,7 @@ async fn login_account_chatgpt_rejected_when_forced_api() -> Result<()> {
 #[tokio::test]
 // Serialize tests that launch the login server since it binds to a fixed port.
 #[serial(login_port)]
-async fn login_account_chatgpt_start() -> Result<()> {
+async fn login_account_chatgpt_start_can_be_cancelled() -> Result<()> {
    let codex_home = TempDir::new()?;
    create_config_toml(codex_home.path(), CreateConfigTomlParams::default())?;

--- a/codex-rs/app-server/tests/suite/v2/config_rpc.rs
+++ b/codex-rs/app-server/tests/suite/v2/config_rpc.rs
@@ -1,6 +1,7 @@
 use anyhow::Result;
 use app_test_support::McpProcess;
 use app_test_support::to_response;
+use codex_app_server_protocol::AskForApproval;
 use codex_app_server_protocol::ConfigBatchWriteParams;
 use codex_app_server_protocol::ConfigEdit;
 use codex_app_server_protocol::ConfigLayerName;
@@ -12,9 +13,12 @@ use codex_app_server_protocol::JSONRPCError;
 use codex_app_server_protocol::JSONRPCResponse;
 use codex_app_server_protocol::MergeStrategy;
 use codex_app_server_protocol::RequestId;
+use codex_app_server_protocol::SandboxMode;
+use codex_app_server_protocol::ToolsV2;
 use codex_app_server_protocol::WriteStatus;
 use pretty_assertions::assert_eq;
 use serde_json::json;
+use std::path::PathBuf;
 use tempfile::TempDir;
 use tokio::time::timeout;

@@ -57,7 +61,7 @@ sandbox_mode = "workspace-write"
        layers,
    } = to_response(resp)?;

-    assert_eq!(config.get("model"), Some(&json!("gpt-user")));
+    assert_eq!(config.model.as_deref(), Some("gpt-user"));
    assert_eq!(
        origins.get("model").expect("origin").name,
        ConfigLayerName::User
@@ -70,6 +74,64 @@ sandbox_mode = "workspace-write"
    Ok(())
 }

+#[tokio::test(flavor = "multi_thread", worker_threads = 2)]
+async fn config_read_includes_tools() -> Result<()> {
+    let codex_home = TempDir::new()?;
+    write_config(
+        &codex_home,
+        r#"
+model = "gpt-user"
+
+[tools]
+web_search = true
+view_image = false
+"#,
+    )?;
+
+    let mut mcp = McpProcess::new(codex_home.path()).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+
+    let request_id = mcp
+        .send_config_read_request(ConfigReadParams {
+            include_layers: true,
+        })
+        .await?;
+    let resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(request_id)),
+    )
+    .await??;
+    let ConfigReadResponse {
+        config,
+        origins,
+        layers,
+    } = to_response(resp)?;
+
+    let tools = config.tools.expect("tools present");
+    assert_eq!(
+        tools,
+        ToolsV2 {
+            web_search: Some(true),
+            view_image: Some(false),
+        }
+    );
+    assert_eq!(
+        origins.get("tools.web_search").expect("origin").name,
+        ConfigLayerName::User
+    );
+    assert_eq!(
+        origins.get("tools.view_image").expect("origin").name,
+        ConfigLayerName::User
+    );
+
+    let layers = layers.expect("layers present");
+    assert_eq!(layers.len(), 2);
+    assert_eq!(layers[0].name, ConfigLayerName::SessionFlags);
+    assert_eq!(layers[1].name, ConfigLayerName::User);
+
+    Ok(())
+}
+
 #[tokio::test(flavor = "multi_thread", worker_threads = 2)]
 async fn config_read_includes_system_layer_and_overrides() -> Result<()> {
    let codex_home = TempDir::new()?;
@@ -123,30 +185,29 @@ writable_roots = ["/system"]
        layers,
    } = to_response(resp)?;

-    assert_eq!(config.get("model"), Some(&json!("gpt-system")));
+    assert_eq!(config.model.as_deref(), Some("gpt-system"));
    assert_eq!(
        origins.get("model").expect("origin").name,
        ConfigLayerName::System
    );

-    assert_eq!(config.get("approval_policy"), Some(&json!("never")));
+    assert_eq!(config.approval_policy, Some(AskForApproval::Never));
    assert_eq!(
        origins.get("approval_policy").expect("origin").name,
        ConfigLayerName::System
    );

-    assert_eq!(config.get("sandbox_mode"), Some(&json!("workspace-write")));
+    assert_eq!(config.sandbox_mode, Some(SandboxMode::WorkspaceWrite));
    assert_eq!(
        origins.get("sandbox_mode").expect("origin").name,
        ConfigLayerName::User
    );

-    assert_eq!(
-        config
-            .get("sandbox_workspace_write")
-            .and_then(|v| v.get("writable_roots")),
-        Some(&json!(["/system"]))
-    );
+    let sandbox = config
+        .sandbox_workspace_write
+        .as_ref()
+        .expect("sandbox workspace write");
+    assert_eq!(sandbox.writable_roots, vec![PathBuf::from("/system")]);
    assert_eq!(
        origins
            .get("sandbox_workspace_write.writable_roots.0")
@@ -155,12 +216,7 @@ writable_roots = ["/system"]
        ConfigLayerName::System
    );

-    assert_eq!(
-        config
-            .get("sandbox_workspace_write")
-            .and_then(|v| v.get("network_access")),
-        Some(&json!(true))
-    );
+    assert!(sandbox.network_access);
    assert_eq!(
        origins
            .get("sandbox_workspace_write.network_access")
@@ -206,7 +262,7 @@ model = "gpt-old"

    let write_id = mcp
        .send_config_value_write_request(ConfigValueWriteParams {
-            file_path: codex_home.path().join("config.toml").display().to_string(),
+            file_path: None,
            key_path: "model".to_string(),
            value: json!("gpt-new"),
            merge_strategy: MergeStrategy::Replace,
@@ -219,8 +275,16 @@ model = "gpt-old"
    )
    .await??;
    let write: ConfigWriteResponse = to_response(write_resp)?;
+    let expected_file_path = codex_home
+        .path()
+        .join("config.toml")
+        .canonicalize()
+        .unwrap()
+        .display()
+        .to_string();

    assert_eq!(write.status, WriteStatus::Ok);
+    assert_eq!(write.file_path, expected_file_path);
    assert!(write.overridden_metadata.is_none());

    let verify_id = mcp
@@ -234,7 +298,7 @@ model = "gpt-old"
    )
    .await??;
    let verify: ConfigReadResponse = to_response(verify_resp)?;
-    assert_eq!(verify.config.get("model"), Some(&json!("gpt-new")));
+    assert_eq!(verify.config.model.as_deref(), Some("gpt-new"));

    Ok(())
 }
@@ -254,7 +318,7 @@ model = "gpt-old"

    let write_id = mcp
        .send_config_value_write_request(ConfigValueWriteParams {
-            file_path: codex_home.path().join("config.toml").display().to_string(),
+            file_path: Some(codex_home.path().join("config.toml").display().to_string()),
            key_path: "model".to_string(),
            value: json!("gpt-new"),
            merge_strategy: MergeStrategy::Replace,
@@ -288,7 +352,7 @@ async fn config_batch_write_applies_multiple_edits() -> Result<()> {

    let batch_id = mcp
        .send_config_batch_write_request(ConfigBatchWriteParams {
-            file_path: codex_home.path().join("config.toml").display().to_string(),
+            file_path: Some(codex_home.path().join("config.toml").display().to_string()),
            edits: vec![
                ConfigEdit {
                    key_path: "sandbox_mode".to_string(),
@@ -314,6 +378,14 @@ async fn config_batch_write_applies_multiple_edits() -> Result<()> {
    .await??;
    let batch_write: ConfigWriteResponse = to_response(batch_resp)?;
    assert_eq!(batch_write.status, WriteStatus::Ok);
+    let expected_file_path = codex_home
+        .path()
+        .join("config.toml")
+        .canonicalize()
+        .unwrap()
+        .display()
+        .to_string();
+    assert_eq!(batch_write.file_path, expected_file_path);

    let read_id = mcp
        .send_config_read_request(ConfigReadParams {
@@ -326,22 +398,14 @@ async fn config_batch_write_applies_multiple_edits() -> Result<()> {
    )
    .await??;
    let read: ConfigReadResponse = to_response(read_resp)?;
-    assert_eq!(
-        read.config.get("sandbox_mode"),
-        Some(&json!("workspace-write"))
-    );
-    assert_eq!(
-        read.config
-            .get("sandbox_workspace_write")
-            .and_then(|v| v.get("writable_roots")),
-        Some(&json!(["/tmp"]))
-    );
-    assert_eq!(
-        read.config
-            .get("sandbox_workspace_write")
-            .and_then(|v| v.get("network_access")),
-        Some(&json!(false))
-    );
+    assert_eq!(read.config.sandbox_mode, Some(SandboxMode::WorkspaceWrite));
+    let sandbox = read
+        .config
+        .sandbox_workspace_write
+        .as_ref()
+        .expect("sandbox workspace write");
+    assert_eq!(sandbox.writable_roots, vec![PathBuf::from("/tmp")]);
+    assert!(!sandbox.network_access);

    Ok(())
 }
--- a/codex-rs/app-server/tests/suite/v2/model_list.rs
+++ b/codex-rs/app-server/tests/suite/v2/model_list.rs
@@ -4,6 +4,7 @@ use anyhow::Result;
 use anyhow::anyhow;
 use app_test_support::McpProcess;
 use app_test_support::to_response;
+use app_test_support::write_models_cache;
 use codex_app_server_protocol::JSONRPCError;
 use codex_app_server_protocol::JSONRPCResponse;
 use codex_app_server_protocol::Model;
@@ -11,7 +12,7 @@ use codex_app_server_protocol::ModelListParams;
 use codex_app_server_protocol::ModelListResponse;
 use codex_app_server_protocol::ReasoningEffortOption;
 use codex_app_server_protocol::RequestId;
-use codex_protocol::config_types::ReasoningEffort;
+use codex_protocol::openai_models::ReasoningEffort;
 use pretty_assertions::assert_eq;
 use tempfile::TempDir;
 use tokio::time::timeout;
@@ -22,6 +23,7 @@ const INVALID_REQUEST_ERROR_CODE: i64 = -32600;
 #[tokio::test]
 async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
    let codex_home = TempDir::new()?;
+    write_models_cache(codex_home.path())?;
    let mut mcp = McpProcess::new(codex_home.path()).await?;

    timeout(DEFAULT_TIMEOUT, mcp.initialize()).await??;
@@ -114,6 +116,37 @@ async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
            default_reasoning_effort: ReasoningEffort::Medium,
            is_default: false,
        },
+        Model {
+            id: "robin".to_string(),
+            model: "robin".to_string(),
+            display_name: "robin".to_string(),
+            description: "Robin".to_string(),
+            supported_reasoning_efforts: vec![
+                ReasoningEffortOption {
+                    reasoning_effort: ReasoningEffort::Low,
+                    description: "Balances speed with some reasoning; useful for straightforward \
+                                   queries and short explanations"
+                        .to_string(),
+                },
+                ReasoningEffortOption {
+                    reasoning_effort: ReasoningEffort::Medium,
+                    description: "Provides a solid balance of reasoning depth and latency for \
+                         general-purpose tasks"
+                        .to_string(),
+                },
+                ReasoningEffortOption {
+                    reasoning_effort: ReasoningEffort::High,
+                    description: "Maximizes reasoning depth for complex or ambiguous problems"
+                        .to_string(),
+                },
+                ReasoningEffortOption {
+                    reasoning_effort: ReasoningEffort::XHigh,
+                    description: "Extra high reasoning for complex problems".to_string(),
+                },
+            ],
+            default_reasoning_effort: ReasoningEffort::Medium,
+            is_default: false,
+        },
        Model {
            id: "gpt-5.1".to_string(),
            model: "gpt-5.1".to_string(),
@@ -151,6 +184,7 @@ async fn list_models_returns_all_models_with_large_limit() -> Result<()> {
 #[tokio::test]
 async fn list_models_pagination_works() -> Result<()> {
    let codex_home = TempDir::new()?;
+    write_models_cache(codex_home.path())?;
    let mut mcp = McpProcess::new(codex_home.path()).await?;

    timeout(DEFAULT_TIMEOUT, mcp.initialize()).await??;
@@ -240,14 +274,37 @@ async fn list_models_pagination_works() -> Result<()> {
    } = to_response::<ModelListResponse>(fourth_response)?;

    assert_eq!(fourth_items.len(), 1);
-    assert_eq!(fourth_items[0].id, "gpt-5.1");
-    assert!(fourth_cursor.is_none());
+    assert_eq!(fourth_items[0].id, "robin");
+    let fifth_cursor = fourth_cursor.ok_or_else(|| anyhow!("cursor for fifth page"))?;
+
+    let fifth_request = mcp
+        .send_list_models_request(ModelListParams {
+            limit: Some(1),
+            cursor: Some(fifth_cursor.clone()),
+        })
+        .await?;
+
+    let fifth_response: JSONRPCResponse = timeout(
+        DEFAULT_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(fifth_request)),
+    )
+    .await??;
+
+    let ModelListResponse {
+        data: fifth_items,
+        next_cursor: fifth_cursor,
+    } = to_response::<ModelListResponse>(fifth_response)?;
+
+    assert_eq!(fifth_items.len(), 1);
+    assert_eq!(fifth_items[0].id, "gpt-5.1");
+    assert!(fifth_cursor.is_none());
    Ok(())
 }

 #[tokio::test]
 async fn list_models_rejects_invalid_cursor() -> Result<()> {
    let codex_home = TempDir::new()?;
+    write_models_cache(codex_home.path())?;
    let mut mcp = McpProcess::new(codex_home.path()).await?;

    timeout(DEFAULT_TIMEOUT, mcp.initialize()).await??;
--- a/codex-rs/app-server/tests/suite/v2/rate_limits.rs
+++ b/codex-rs/app-server/tests/suite/v2/rate_limits.rs
@@ -11,6 +11,7 @@ use codex_app_server_protocol::RateLimitSnapshot;
 use codex_app_server_protocol::RateLimitWindow;
 use codex_app_server_protocol::RequestId;
 use codex_core::auth::AuthCredentialsStoreMode;
+use codex_protocol::account::PlanType as AccountPlanType;
 use pretty_assertions::assert_eq;
 use serde_json::json;
 use std::path::Path;
@@ -153,6 +154,7 @@ async fn get_account_rate_limits_returns_snapshot() -> Result<()> {
                resets_at: Some(secondary_reset_timestamp),
            }),
            credits: None,
+            plan_type: Some(AccountPlanType::Pro),
        },
    };
    assert_eq!(received, expected);
--- a/codex-rs/app-server/tests/suite/v2/review.rs
+++ b/codex-rs/app-server/tests/suite/v2/review.rs
@@ -9,12 +9,13 @@ use codex_app_server_protocol::JSONRPCError;
 use codex_app_server_protocol::JSONRPCNotification;
 use codex_app_server_protocol::JSONRPCResponse;
 use codex_app_server_protocol::RequestId;
+use codex_app_server_protocol::ReviewDelivery;
 use codex_app_server_protocol::ReviewStartParams;
+use codex_app_server_protocol::ReviewStartResponse;
 use codex_app_server_protocol::ReviewTarget;
 use codex_app_server_protocol::ThreadItem;
 use codex_app_server_protocol::ThreadStartParams;
 use codex_app_server_protocol::ThreadStartResponse;
-use codex_app_server_protocol::TurnStartResponse;
 use codex_app_server_protocol::TurnStatus;
 use serde_json::json;
 use tempfile::TempDir;
@@ -59,7 +60,7 @@ async fn review_start_runs_review_turn_and_emits_code_review_item() -> Result<()
    let review_req = mcp
        .send_review_start_request(ReviewStartParams {
            thread_id: thread_id.clone(),
-            append_to_original_thread: true,
+            delivery: Some(ReviewDelivery::Inline),
            target: ReviewTarget::Commit {
                sha: "1234567deadbeef".to_string(),
                title: Some("Tidy UI colors".to_string()),
@@ -71,43 +72,43 @@ async fn review_start_runs_review_turn_and_emits_code_review_item() -> Result<()
        mcp.read_stream_until_response_message(RequestId::Integer(review_req)),
    )
    .await??;
-    let TurnStartResponse { turn } = to_response::<TurnStartResponse>(review_resp)?;
+    let ReviewStartResponse {
+        turn,
+        review_thread_id,
+    } = to_response::<ReviewStartResponse>(review_resp)?;
+    assert_eq!(review_thread_id, thread_id.clone());
    let turn_id = turn.id.clone();
    assert_eq!(turn.status, TurnStatus::InProgress);
-    assert_eq!(turn.items.len(), 1);
-    match &turn.items[0] {
-        ThreadItem::UserMessage { content, .. } => {
-            assert_eq!(content.len(), 1);
-            assert!(matches!(
-                &content[0],
-                codex_app_server_protocol::UserInput::Text { .. }
-            ));
-        }
-        other => panic!("expected user message, got {other:?}"),
-    }

-    let _started: JSONRPCNotification = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_notification_message("turn/started"),
-    )
-    .await??;
-    let item_started: JSONRPCNotification = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_notification_message("item/started"),
-    )
-    .await??;
-    let started: ItemStartedNotification =
-        serde_json::from_value(item_started.params.expect("params must be present"))?;
-    match started.item {
-        ThreadItem::CodeReview { id, review } => {
-            assert_eq!(id, turn_id);
-            assert_eq!(review, "commit 1234567");
+    // Confirm we see the EnteredReviewMode marker on the main thread.
+    let mut saw_entered_review_mode = false;
+    for _ in 0..10 {
+        let item_started: JSONRPCNotification = timeout(
+            DEFAULT_READ_TIMEOUT,
+            mcp.read_stream_until_notification_message("item/started"),
+        )
+        .await??;
+        let started: ItemStartedNotification =
+            serde_json::from_value(item_started.params.expect("params must be present"))?;
+        match started.item {
+            ThreadItem::EnteredReviewMode { id, review } => {
+                assert_eq!(id, turn_id);
+                assert_eq!(review, "commit 1234567: Tidy UI colors");
+                saw_entered_review_mode = true;
+                break;
+            }
+            _ => continue,
        }
-        other => panic!("expected code review item, got {other:?}"),
    }
+    assert!(
+        saw_entered_review_mode,
+        "did not observe enteredReviewMode item"
+    );

+    // Confirm we see the ExitedReviewMode marker (with review text)
+    // on the same turn. Ignore any other items the stream surfaces.
    let mut review_body: Option<String> = None;
-    for _ in 0..5 {
+    for _ in 0..10 {
        let review_notif: JSONRPCNotification = timeout(
            DEFAULT_READ_TIMEOUT,
            mcp.read_stream_until_notification_message("item/completed"),
@@ -116,13 +117,12 @@ async fn review_start_runs_review_turn_and_emits_code_review_item() -> Result<()
        let completed: ItemCompletedNotification =
            serde_json::from_value(review_notif.params.expect("params must be present"))?;
        match completed.item {
-            ThreadItem::CodeReview { id, review } => {
+            ThreadItem::ExitedReviewMode { id, review } => {
                assert_eq!(id, turn_id);
                review_body = Some(review);
                break;
            }
-            ThreadItem::UserMessage { .. } => continue,
-            other => panic!("unexpected item/completed payload: {other:?}"),
+            _ => continue,
        }
    }

@@ -146,7 +146,7 @@ async fn review_start_rejects_empty_base_branch() -> Result<()> {
    let request_id = mcp
        .send_review_start_request(ReviewStartParams {
            thread_id,
-            append_to_original_thread: true,
+            delivery: Some(ReviewDelivery::Inline),
            target: ReviewTarget::BaseBranch {
                branch: "   ".to_string(),
            },
@@ -167,6 +167,56 @@ async fn review_start_rejects_empty_base_branch() -> Result<()> {
    Ok(())
 }

+#[tokio::test]
+async fn review_start_with_detached_delivery_returns_new_thread_id() -> Result<()> {
+    let review_payload = json!({
+        "findings": [],
+        "overall_correctness": "ok",
+        "overall_explanation": "detached review",
+        "overall_confidence_score": 0.5
+    })
+    .to_string();
+    let responses = vec![create_final_assistant_message_sse_response(
+        &review_payload,
+    )?];
+    let server = create_mock_chat_completions_server_unchecked(responses).await;
+
+    let codex_home = TempDir::new()?;
+    create_config_toml(codex_home.path(), &server.uri())?;
+
+    let mut mcp = McpProcess::new(codex_home.path()).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+
+    let thread_id = start_default_thread(&mut mcp).await?;
+
+    let review_req = mcp
+        .send_review_start_request(ReviewStartParams {
+            thread_id: thread_id.clone(),
+            delivery: Some(ReviewDelivery::Detached),
+            target: ReviewTarget::Custom {
+                instructions: "detached review".to_string(),
+            },
+        })
+        .await?;
+    let review_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(review_req)),
+    )
+    .await??;
+    let ReviewStartResponse {
+        turn,
+        review_thread_id,
+    } = to_response::<ReviewStartResponse>(review_resp)?;
+
+    assert_eq!(turn.status, TurnStatus::InProgress);
+    assert_ne!(
+        review_thread_id, thread_id,
+        "detached review should run on a different thread"
+    );
+
+    Ok(())
+}
+
 #[tokio::test]
 async fn review_start_rejects_empty_commit_sha() -> Result<()> {
    let server = create_mock_chat_completions_server_unchecked(vec![]).await;
@@ -180,7 +230,7 @@ async fn review_start_rejects_empty_commit_sha() -> Result<()> {
    let request_id = mcp
        .send_review_start_request(ReviewStartParams {
            thread_id,
-            append_to_original_thread: true,
+            delivery: Some(ReviewDelivery::Inline),
            target: ReviewTarget::Commit {
                sha: "\t".to_string(),
                title: None,
@@ -215,7 +265,7 @@ async fn review_start_rejects_empty_custom_instructions() -> Result<()> {
    let request_id = mcp
        .send_review_start_request(ReviewStartParams {
            thread_id,
-            append_to_original_thread: true,
+            delivery: Some(ReviewDelivery::Inline),
            target: ReviewTarget::Custom {
                instructions: "\n\n".to_string(),
            },
--- a/codex-rs/app-server/tests/suite/v2/thread_list.rs
+++ b/codex-rs/app-server/tests/suite/v2/thread_list.rs
@@ -6,37 +6,96 @@ use codex_app_server_protocol::GitInfo as ApiGitInfo;
 use codex_app_server_protocol::JSONRPCResponse;
 use codex_app_server_protocol::RequestId;
 use codex_app_server_protocol::SessionSource;
-use codex_app_server_protocol::ThreadListParams;
 use codex_app_server_protocol::ThreadListResponse;
 use codex_protocol::protocol::GitInfo as CoreGitInfo;
+use std::path::Path;
 use std::path::PathBuf;
 use tempfile::TempDir;
 use tokio::time::timeout;

 const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);

+async fn init_mcp(codex_home: &Path) -> Result<McpProcess> {
+    let mut mcp = McpProcess::new(codex_home).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+    Ok(mcp)
+}
+
+async fn list_threads(
+    mcp: &mut McpProcess,
+    cursor: Option<String>,
+    limit: Option<u32>,
+    providers: Option<Vec<String>>,
+) -> Result<ThreadListResponse> {
+    let request_id = mcp
+        .send_thread_list_request(codex_app_server_protocol::ThreadListParams {
+            cursor,
+            limit,
+            model_providers: providers,
+        })
+        .await?;
+    let resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(request_id)),
+    )
+    .await??;
+    to_response::<ThreadListResponse>(resp)
+}
+
+fn create_fake_rollouts<F, G>(
+    codex_home: &Path,
+    count: usize,
+    provider_for_index: F,
+    timestamp_for_index: G,
+    preview: &str,
+) -> Result<Vec<String>>
+where
+    F: Fn(usize) -> &'static str,
+    G: Fn(usize) -> (String, String),
+{
+    let mut ids = Vec::with_capacity(count);
+    for i in 0..count {
+        let (ts_file, ts_rfc) = timestamp_for_index(i);
+        ids.push(create_fake_rollout(
+            codex_home,
+            &ts_file,
+            &ts_rfc,
+            preview,
+            Some(provider_for_index(i)),
+            None,
+        )?);
+    }
+    Ok(ids)
+}
+
+fn timestamp_at(
+    year: i32,
+    month: u32,
+    day: u32,
+    hour: u32,
+    minute: u32,
+    second: u32,
+) -> (String, String) {
+    (
+        format!("{year:04}-{month:02}-{day:02}T{hour:02}-{minute:02}-{second:02}"),
+        format!("{year:04}-{month:02}-{day:02}T{hour:02}:{minute:02}:{second:02}Z"),
+    )
+}
+
 #[tokio::test]
 async fn thread_list_basic_empty() -> Result<()> {
    let codex_home = TempDir::new()?;
    create_minimal_config(codex_home.path())?;

-    let mut mcp = McpProcess::new(codex_home.path()).await?;
-    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+    let mut mcp = init_mcp(codex_home.path()).await?;

-    // List threads in an empty CODEX_HOME; should return an empty page with nextCursor: null.
-    let list_id = mcp
-        .send_thread_list_request(ThreadListParams {
-            cursor: None,
-            limit: Some(10),
-            model_providers: Some(vec!["mock_provider".to_string()]),
-        })
-        .await?;
-    let list_resp: JSONRPCResponse = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_response_message(RequestId::Integer(list_id)),
+    let ThreadListResponse { data, next_cursor } = list_threads(
+        &mut mcp,
+        None,
+        Some(10),
+        Some(vec!["mock_provider".to_string()]),
    )
-    .await??;
-    let ThreadListResponse { data, next_cursor } = to_response::<ThreadListResponse>(list_resp)?;
+    .await?;
    assert!(data.is_empty());
    assert_eq!(next_cursor, None);

@@ -86,26 +145,19 @@ async fn thread_list_pagination_next_cursor_none_on_last_page() -> Result<()> {
        None,
    )?;

-    let mut mcp = McpProcess::new(codex_home.path()).await?;
-    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+    let mut mcp = init_mcp(codex_home.path()).await?;

    // Page 1: limit 2 → expect next_cursor Some.
-    let page1_id = mcp
-        .send_thread_list_request(ThreadListParams {
-            cursor: None,
-            limit: Some(2),
-            model_providers: Some(vec!["mock_provider".to_string()]),
-        })
-        .await?;
-    let page1_resp: JSONRPCResponse = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_response_message(RequestId::Integer(page1_id)),
-    )
-    .await??;
    let ThreadListResponse {
        data: data1,
        next_cursor: cursor1,
-    } = to_response::<ThreadListResponse>(page1_resp)?;
+    } = list_threads(
+        &mut mcp,
+        None,
+        Some(2),
+        Some(vec!["mock_provider".to_string()]),
+    )
+    .await?;
    assert_eq!(data1.len(), 2);
    for thread in &data1 {
        assert_eq!(thread.preview, "Hello");
@@ -119,22 +171,16 @@ async fn thread_list_pagination_next_cursor_none_on_last_page() -> Result<()> {
    let cursor1 = cursor1.expect("expected nextCursor on first page");

    // Page 2: with cursor → expect next_cursor None when no more results.
-    let page2_id = mcp
-        .send_thread_list_request(ThreadListParams {
-            cursor: Some(cursor1),
-            limit: Some(2),
-            model_providers: Some(vec!["mock_provider".to_string()]),
-        })
-        .await?;
-    let page2_resp: JSONRPCResponse = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_response_message(RequestId::Integer(page2_id)),
-    )
-    .await??;
    let ThreadListResponse {
        data: data2,
        next_cursor: cursor2,
-    } = to_response::<ThreadListResponse>(page2_resp)?;
+    } = list_threads(
+        &mut mcp,
+        Some(cursor1),
+        Some(2),
+        Some(vec!["mock_provider".to_string()]),
+    )
+    .await?;
    assert!(data2.len() <= 2);
    for thread in &data2 {
        assert_eq!(thread.preview, "Hello");
@@ -173,23 +219,16 @@ async fn thread_list_respects_provider_filter() -> Result<()> {
        None,
    )?;

-    let mut mcp = McpProcess::new(codex_home.path()).await?;
-    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+    let mut mcp = init_mcp(codex_home.path()).await?;

    // Filter to only other_provider; expect 1 item, nextCursor None.
-    let list_id = mcp
-        .send_thread_list_request(ThreadListParams {
-            cursor: None,
-            limit: Some(10),
-            model_providers: Some(vec!["other_provider".to_string()]),
-        })
-        .await?;
-    let resp: JSONRPCResponse = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_response_message(RequestId::Integer(list_id)),
+    let ThreadListResponse { data, next_cursor } = list_threads(
+        &mut mcp,
+        None,
+        Some(10),
+        Some(vec!["other_provider".to_string()]),
    )
-    .await??;
-    let ThreadListResponse { data, next_cursor } = to_response::<ThreadListResponse>(resp)?;
+    .await?;
    assert_eq!(data.len(), 1);
    assert_eq!(next_cursor, None);
    let thread = &data[0];
@@ -205,6 +244,146 @@ async fn thread_list_respects_provider_filter() -> Result<()> {
    Ok(())
 }

+#[tokio::test]
+async fn thread_list_fetches_until_limit_or_exhausted() -> Result<()> {
+    let codex_home = TempDir::new()?;
+    create_minimal_config(codex_home.path())?;
+
+    // Newest 16 conversations belong to a different provider; the older 8 are the
+    // only ones that match the filter. We request 8 so the server must keep
+    // paging past the first two pages to reach the desired count.
+    create_fake_rollouts(
+        codex_home.path(),
+        24,
+        |i| {
+            if i < 16 {
+                "skip_provider"
+            } else {
+                "target_provider"
+            }
+        },
+        |i| timestamp_at(2025, 3, 30 - i as u32, 12, 0, 0),
+        "Hello",
+    )?;
+
+    let mut mcp = init_mcp(codex_home.path()).await?;
+
+    // Request 8 threads for the target provider; the matches only start on the
+    // third page so we rely on pagination to reach the limit.
+    let ThreadListResponse { data, next_cursor } = list_threads(
+        &mut mcp,
+        None,
+        Some(8),
+        Some(vec!["target_provider".to_string()]),
+    )
+    .await?;
+    assert_eq!(
+        data.len(),
+        8,
+        "should keep paging until the requested count is filled"
+    );
+    assert!(
+        data.iter()
+            .all(|thread| thread.model_provider == "target_provider"),
+        "all returned threads must match the requested provider"
+    );
+    assert_eq!(
+        next_cursor, None,
+        "once the requested count is satisfied on the final page, nextCursor should be None"
+    );
+
+    Ok(())
+}
+
+#[tokio::test]
+async fn thread_list_enforces_max_limit() -> Result<()> {
+    let codex_home = TempDir::new()?;
+    create_minimal_config(codex_home.path())?;
+
+    create_fake_rollouts(
+        codex_home.path(),
+        105,
+        |_| "mock_provider",
+        |i| {
+            let month = 5 + (i / 28);
+            let day = (i % 28) + 1;
+            timestamp_at(2025, month as u32, day as u32, 0, 0, 0)
+        },
+        "Hello",
+    )?;
+
+    let mut mcp = init_mcp(codex_home.path()).await?;
+
+    let ThreadListResponse { data, next_cursor } = list_threads(
+        &mut mcp,
+        None,
+        Some(200),
+        Some(vec!["mock_provider".to_string()]),
+    )
+    .await?;
+    assert_eq!(
+        data.len(),
+        100,
+        "limit should be clamped to the maximum page size"
+    );
+    assert!(
+        next_cursor.is_some(),
+        "when more than the maximum exist, nextCursor should continue pagination"
+    );
+
+    Ok(())
+}
+
+#[tokio::test]
+async fn thread_list_stops_when_not_enough_filtered_results_exist() -> Result<()> {
+    let codex_home = TempDir::new()?;
+    create_minimal_config(codex_home.path())?;
+
+    // Only the last 7 conversations match the provider filter; we ask for 10 to
+    // ensure the server exhausts pagination without looping forever.
+    create_fake_rollouts(
+        codex_home.path(),
+        22,
+        |i| {
+            if i < 15 {
+                "skip_provider"
+            } else {
+                "target_provider"
+            }
+        },
+        |i| timestamp_at(2025, 4, 28 - i as u32, 8, 0, 0),
+        "Hello",
+    )?;
+
+    let mut mcp = init_mcp(codex_home.path()).await?;
+
+    // Request more threads than exist after filtering; expect all matches to be
+    // returned with nextCursor None.
+    let ThreadListResponse { data, next_cursor } = list_threads(
+        &mut mcp,
+        None,
+        Some(10),
+        Some(vec!["target_provider".to_string()]),
+    )
+    .await?;
+    assert_eq!(
+        data.len(),
+        7,
+        "all available filtered threads should be returned"
+    );
+    assert!(
+        data.iter()
+            .all(|thread| thread.model_provider == "target_provider"),
+        "results should still respect the provider filter"
+    );
+    assert_eq!(
+        next_cursor, None,
+        "when results are exhausted before reaching the limit, nextCursor should be None"
+    );
+
+    Ok(())
+}
+
 #[tokio::test]
 async fn thread_list_includes_git_info() -> Result<()> {
    let codex_home = TempDir::new()?;
@@ -224,22 +403,15 @@ async fn thread_list_includes_git_info() -> Result<()> {
        Some(git_info),
    )?;

-    let mut mcp = McpProcess::new(codex_home.path()).await?;
-    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+    let mut mcp = init_mcp(codex_home.path()).await?;

-    let list_id = mcp
-        .send_thread_list_request(ThreadListParams {
-            cursor: None,
-            limit: Some(10),
-            model_providers: Some(vec!["mock_provider".to_string()]),
-        })
-        .await?;
-    let resp: JSONRPCResponse = timeout(
-        DEFAULT_READ_TIMEOUT,
-        mcp.read_stream_until_response_message(RequestId::Integer(list_id)),
+    let ThreadListResponse { data, .. } = list_threads(
+        &mut mcp,
+        None,
+        Some(10),
+        Some(vec!["mock_provider".to_string()]),
    )
-    .await??;
-    let ThreadListResponse { data, .. } = to_response::<ThreadListResponse>(resp)?;
+    .await?;
    let thread = data
        .iter()
        .find(|t| t.id == conversation_id)
--- a/codex-rs/app-server/tests/suite/v2/turn_start.rs
+++ b/codex-rs/app-server/tests/suite/v2/turn_start.rs
@@ -11,6 +11,7 @@ use app_test_support::to_response;
 use codex_app_server_protocol::ApprovalDecision;
 use codex_app_server_protocol::CommandExecutionRequestApprovalResponse;
 use codex_app_server_protocol::CommandExecutionStatus;
+use codex_app_server_protocol::FileChangeOutputDeltaNotification;
 use codex_app_server_protocol::FileChangeRequestApprovalResponse;
 use codex_app_server_protocol::ItemCompletedNotification;
 use codex_app_server_protocol::ItemStartedNotification;
@@ -29,8 +30,8 @@ use codex_app_server_protocol::TurnStartResponse;
 use codex_app_server_protocol::TurnStartedNotification;
 use codex_app_server_protocol::TurnStatus;
 use codex_app_server_protocol::UserInput as V2UserInput;
-use codex_core::protocol_config_types::ReasoningEffort;
 use codex_core::protocol_config_types::ReasoningSummary;
+use codex_protocol::openai_models::ReasoningEffort;
 use core_test_support::skip_if_no_network;
 use pretty_assertions::assert_eq;
 use std::path::Path;
@@ -426,7 +427,6 @@ async fn turn_start_exec_approval_decline_v2() -> Result<()> {
        request_id,
        serde_json::to_value(CommandExecutionRequestApprovalResponse {
            decision: ApprovalDecision::Decline,
-            accept_settings: None,
        })?,
    )
    .await?;
@@ -725,6 +725,26 @@ async fn turn_start_file_change_approval_v2() -> Result<()> {
    )
    .await?;

+    let output_delta_notif = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_notification_message("item/fileChange/outputDelta"),
+    )
+    .await??;
+    let output_delta: FileChangeOutputDeltaNotification = serde_json::from_value(
+        output_delta_notif
+            .params
+            .clone()
+            .expect("item/fileChange/outputDelta params"),
+    )?;
+    assert_eq!(output_delta.thread_id, thread.id);
+    assert_eq!(output_delta.turn_id, turn.id);
+    assert_eq!(output_delta.item_id, "patch-call");
+    assert!(
+        !output_delta.delta.is_empty(),
+        "expected delta to be non-empty, got: {}",
+        output_delta.delta
+    );
+
    let completed_file_change = timeout(DEFAULT_READ_TIMEOUT, async {
        loop {
            let completed_notif = mcp
--- a/codex-rs/apply-patch/src/lib.rs
+++ b/codex-rs/apply-patch/src/lib.rs
@@ -112,7 +112,7 @@ fn classify_shell_name(shell: &str) -> Option<String> {

 fn classify_shell(shell: &str, flag: &str) -> Option<ApplyPatchShell> {
    classify_shell_name(shell).and_then(|name| match name.as_str() {
-        "bash" | "zsh" | "sh" if flag == "-lc" => Some(ApplyPatchShell::Unix),
+        "bash" | "zsh" | "sh" if matches!(flag, "-lc" | "-c") => Some(ApplyPatchShell::Unix),
        "pwsh" | "powershell" if flag.eq_ignore_ascii_case("-command") => {
            Some(ApplyPatchShell::PowerShell)
        }
@@ -699,13 +699,7 @@ fn derive_new_contents_from_chunks(
        }
    };

-    let mut original_lines: Vec<String> = original_contents.split('\n').map(String::from).collect();
-
-    // Drop the trailing empty element that results from the final newline so
-    // that line counts match the behaviour of standard `diff`.
-    if original_lines.last().is_some_and(String::is_empty) {
-        original_lines.pop();
-    }
+    let original_lines: Vec<String> = build_lines_from_contents(&original_contents);

    let replacements = compute_replacements(&original_lines, path, chunks)?;
    let new_lines = apply_replacements(original_lines, &replacements);
@@ -713,13 +707,67 @@ fn derive_new_contents_from_chunks(
    if !new_lines.last().is_some_and(String::is_empty) {
        new_lines.push(String::new());
    }
-    let new_contents = new_lines.join("\n");
+    let new_contents = build_contents_from_lines(&original_contents, &new_lines);
    Ok(AppliedPatch {
        original_contents,
        new_contents,
    })
 }

+// TODO(dylan-hurd-oai): I think we can migrate to just use `contents.lines()`
+// across all platforms.
+fn build_lines_from_contents(contents: &str) -> Vec<String> {
+    if cfg!(windows) {
+        contents.lines().map(String::from).collect()
+    } else {
+        let mut lines: Vec<String> = contents.split('\n').map(String::from).collect();
+
+        // Drop the trailing empty element that results from the final newline so
+        // that line counts match the behaviour of standard `diff`.
+        if lines.last().is_some_and(String::is_empty) {
+            lines.pop();
+        }
+
+        lines
+    }
+}
+
+fn build_contents_from_lines(original_contents: &str, lines: &[String]) -> String {
+    if cfg!(windows) {
+        // for now, only compute this if we're on Windows.
+        let uses_crlf = contents_uses_crlf(original_contents);
+        if uses_crlf {
+            lines.join("\r\n")
+        } else {
+            lines.join("\n")
+        }
+    } else {
+        lines.join("\n")
+    }
+}
+
+/// Detects whether the source file uses Windows CRLF line endings consistently.
+/// We only consider a file CRLF-formatted if every newline is part of a
+/// CRLF sequence. This avoids rewriting an LF-formatted file that merely
+/// contains embedded sequences of "\r\n".
+///
+/// Returns `true` if the file uses CRLF line endings, `false` otherwise.
+fn contents_uses_crlf(contents: &str) -> bool {
+    let bytes = contents.as_bytes();
+    let mut n_newlines = 0usize;
+    let mut n_crlf = 0usize;
+    for i in 0..bytes.len() {
+        if bytes[i] == b'\n' {
+            n_newlines += 1;
+            if i > 0 && bytes[i - 1] == b'\r' {
+                n_crlf += 1;
+            }
+        }
+    }
+
+    n_newlines > 0 && n_crlf == n_newlines
+}
+
 /// Compute a list of replacements needed to transform `original_lines` into the
 /// new lines, given the patch `chunks`. Each replacement is returned as
 /// `(start_index, old_len, new_lines)`.
@@ -1049,6 +1097,13 @@ mod tests {
        assert_match(&heredoc_script(""), None);
    }

+    #[test]
+    fn test_heredoc_non_login_shell() {
+        let script = heredoc_script("");
+        let args = strs_to_strings(&["bash", "-c", &script]);
+        assert_match_args(args, None);
+    }
+
    #[test]
    fn test_heredoc_applypatch() {
        let args = strs_to_strings(&[
@@ -1359,6 +1414,72 @@ PATCH"#,
        assert_eq!(contents, "a\nB\nc\nd\nE\nf\ng\n");
    }

+    /// Ensure CRLF line endings are preserved for updated files on Windows‑style inputs.
+    #[cfg(windows)]
+    #[test]
+    fn test_preserve_crlf_line_endings_on_update() {
+        let dir = tempdir().unwrap();
+        let path = dir.path().join("crlf.txt");
+
+        // Original file uses CRLF (\r\n) endings.
+        std::fs::write(&path, b"a\r\nb\r\nc\r\n").unwrap();
+
+        // Replace `b` -> `B` and append `d`.
+        let patch = wrap_patch(&format!(
+            r#"*** Update File: {}
+@@
+ a
+-b
+B
+@@
+ c
+d
+*** End of File"#,
+            path.display()
+        ));
+
+        let mut stdout = Vec::new();
+        let mut stderr = Vec::new();
+        apply_patch(&patch, &mut stdout, &mut stderr).unwrap();
+
+        let out = std::fs::read(&path).unwrap();
+        // Expect all CRLF endings; count occurrences of CRLF and ensure there are 4 lines.
+        let content = String::from_utf8_lossy(&out);
+        assert!(content.contains("\r\n"));
+        // No bare LF occurrences immediately preceding a non-CR: the text should not contain "a\nb".
+        assert!(!content.contains("a\nb"));
+        // Validate exact content sequence with CRLF delimiters.
+        assert_eq!(content, "a\r\nB\r\nc\r\nd\r\n");
+    }
+
+    /// Ensure CRLF inputs with embedded carriage returns in the content are preserved.
+    #[cfg(windows)]
+    #[test]
+    fn test_preserve_crlf_embedded_carriage_returns_on_append() {
+        let dir = tempdir().unwrap();
+        let path = dir.path().join("crlf_cr_content.txt");
+
+        // Original file: first line has a literal '\r' in the content before the CRLF terminator.
+        std::fs::write(&path, b"foo\r\r\nbar\r\n").unwrap();
+
+        // Append a new line without modifying existing ones.
+        let patch = wrap_patch(&format!(
+            r#"*** Update File: {}
+@@
+BAZ
+*** End of File"#,
+            path.display()
+        ));
+
+        let mut stdout = Vec::new();
+        let mut stderr = Vec::new();
+        apply_patch(&patch, &mut stdout, &mut stderr).unwrap();
+
+        let out = std::fs::read(&path).unwrap();
+        // CRLF endings must be preserved and the extra CR in "foo\r\r" must not be collapsed.
+        assert_eq!(out.as_slice(), b"foo\r\r\nbar\r\nBAZ\r\n");
+    }
+
    #[test]
    fn test_pure_addition_chunk_followed_by_removal() {
        let dir = tempdir().unwrap();
@@ -1544,6 +1665,37 @@ PATCH"#,
        assert_eq!(expected, diff);
    }

+    /// For LF-only inputs with a trailing newline ensure that the helper used
+    /// on Windows-style builds drops the synthetic trailing empty element so
+    /// replacements behave like standard `diff` line numbering.
+    #[test]
+    fn test_derive_new_contents_lf_trailing_newline() {
+        let dir = tempdir().unwrap();
+        let path = dir.path().join("lf_trailing_newline.txt");
+        fs::write(&path, "foo\nbar\n").unwrap();
+
+        let patch = wrap_patch(&format!(
+            r#"*** Update File: {}
+@@
+ foo
+-bar
+BAR
+"#,
+            path.display()
+        ));
+
+        let patch = parse_patch(&patch).unwrap();
+        let chunks = match patch.hunks.as_slice() {
+            [Hunk::UpdateFile { chunks, .. }] => chunks,
+            _ => panic!("Expected a single UpdateFile hunk"),
+        };
+
+        let AppliedPatch { new_contents, .. } =
+            derive_new_contents_from_chunks(&path, chunks).unwrap();
+
+        assert_eq!(new_contents, "foo\nBAR\n");
+    }
+
    #[test]
    fn test_unified_diff_insert_at_eof() {
        // Insert a new line at end‑of‑file.
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/.gitattributes
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/.gitattributes
@@ -0,0 +1 @@
+** text eol=lf
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/001_add_file/expected/bar.md
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/001_add_file/expected/bar.md
@@ -0,0 +1 @@
+This is a new file
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/001_add_file/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/001_add_file/patch.txt
@@ -0,0 +1,4 @@
+*** Begin Patch
+*** Add File: bar.md
+This is a new file
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/expected/modify.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/expected/modify.txt
@@ -0,0 +1,2 @@
+line1
+changed
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/expected/nested/new.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/expected/nested/new.txt
@@ -0,0 +1 @@
+created
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/input/delete.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/input/delete.txt
@@ -0,0 +1 @@
+obsolete
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/input/modify.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/input/modify.txt
@@ -0,0 +1,2 @@
+line1
+line2
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/002_multiple_operations/patch.txt
@@ -0,0 +1,9 @@
+*** Begin Patch
+*** Add File: nested/new.txt
+created
+*** Delete File: delete.txt
+*** Update File: modify.txt
+@@
+-line2
+changed
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/003_multiple_chunks/expected/multi.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/003_multiple_chunks/expected/multi.txt
@@ -0,0 +1,4 @@
+line1
+changed2
+line3
+changed4
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/003_multiple_chunks/input/multi.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/003_multiple_chunks/input/multi.txt
@@ -0,0 +1,4 @@
+line1
+line2
+line3
+line4
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/003_multiple_chunks/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/003_multiple_chunks/patch.txt
@@ -0,0 +1,9 @@
+*** Begin Patch
+*** Update File: multi.txt
+@@
+-line2
+changed2
+@@
+-line4
+changed4
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/expected/old/other.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/expected/old/other.txt
@@ -0,0 +1 @@
+unrelated file
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/expected/renamed/dir/name.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/expected/renamed/dir/name.txt
@@ -0,0 +1 @@
+new content
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/input/old/name.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/input/old/name.txt
@@ -0,0 +1 @@
+old content
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/input/old/other.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/input/old/other.txt
@@ -0,0 +1 @@
+unrelated file
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/004_move_to_new_directory/patch.txt
@@ -0,0 +1,7 @@
+*** Begin Patch
+*** Update File: old/name.txt
+*** Move to: renamed/dir/name.txt
+@@
+-old content
+new content
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/005_rejects_empty_patch/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/005_rejects_empty_patch/patch.txt
@@ -0,0 +1,2 @@
+*** Begin Patch
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/006_rejects_missing_context/expected/modify.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/006_rejects_missing_context/expected/modify.txt
@@ -0,0 +1,2 @@
+line1
+line2
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/006_rejects_missing_context/input/modify.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/006_rejects_missing_context/input/modify.txt
@@ -0,0 +1,2 @@
+line1
+line2
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/006_rejects_missing_context/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/006_rejects_missing_context/patch.txt
@@ -0,0 +1,6 @@
+*** Begin Patch
+*** Update File: modify.txt
+@@
+-missing
+changed
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/007_rejects_missing_file_delete/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/007_rejects_missing_file_delete/patch.txt
@@ -0,0 +1,3 @@
+*** Begin Patch
+*** Delete File: missing.txt
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/008_rejects_empty_update_hunk/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/008_rejects_empty_update_hunk/patch.txt
@@ -0,0 +1,3 @@
+*** Begin Patch
+*** Update File: foo.txt
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/009_requires_existing_file_for_update/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/009_requires_existing_file_for_update/patch.txt
@@ -0,0 +1,6 @@
+*** Begin Patch
+*** Update File: missing.txt
+@@
+-old
+new
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/expected/old/other.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/expected/old/other.txt
@@ -0,0 +1 @@
+unrelated file
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/expected/renamed/dir/name.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/expected/renamed/dir/name.txt
@@ -0,0 +1 @@
+new
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/input/old/name.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/input/old/name.txt
@@ -0,0 +1 @@
+from
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/input/old/other.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/input/old/other.txt
@@ -0,0 +1 @@
+unrelated file
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/input/renamed/dir/name.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/input/renamed/dir/name.txt
@@ -0,0 +1 @@
+existing
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/010_move_overwrites_existing_destination/patch.txt
@@ -0,0 +1,7 @@
+*** Begin Patch
+*** Update File: old/name.txt
+*** Move to: renamed/dir/name.txt
+@@
+-from
+new
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/011_add_overwrites_existing_file/expected/duplicate.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/011_add_overwrites_existing_file/expected/duplicate.txt
@@ -0,0 +1 @@
+new content
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/011_add_overwrites_existing_file/input/duplicate.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/011_add_overwrites_existing_file/input/duplicate.txt
@@ -0,0 +1 @@
+old content
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/011_add_overwrites_existing_file/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/011_add_overwrites_existing_file/patch.txt
@@ -0,0 +1,4 @@
+*** Begin Patch
+*** Add File: duplicate.txt
+new content
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/012_delete_directory_fails/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/012_delete_directory_fails/patch.txt
@@ -0,0 +1,3 @@
+*** Begin Patch
+*** Delete File: dir
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/013_rejects_invalid_hunk_header/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/013_rejects_invalid_hunk_header/patch.txt
@@ -0,0 +1,3 @@
+*** Begin Patch
+*** Frobnicate File: foo
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/014_update_file_appends_trailing_newline/expected/no_newline.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/014_update_file_appends_trailing_newline/expected/no_newline.txt
@@ -0,0 +1,2 @@
+first line
+second line
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/014_update_file_appends_trailing_newline/input/no_newline.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/014_update_file_appends_trailing_newline/input/no_newline.txt
@@ -0,0 +1 @@
+no newline at end
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/014_update_file_appends_trailing_newline/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/014_update_file_appends_trailing_newline/patch.txt
@@ -0,0 +1,7 @@
+*** Begin Patch
+*** Update File: no_newline.txt
+@@
+-no newline at end
+first line
+second line
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/015_failure_after_partial_success_leaves_changes/expected/created.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/015_failure_after_partial_success_leaves_changes/expected/created.txt
@@ -0,0 +1 @@
+hello
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/015_failure_after_partial_success_leaves_changes/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/015_failure_after_partial_success_leaves_changes/patch.txt
@@ -0,0 +1,8 @@
+*** Begin Patch
+*** Add File: created.txt
+hello
+*** Update File: missing.txt
+@@
+-old
+new
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/016_pure_addition_update_chunk/expected/input.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/016_pure_addition_update_chunk/expected/input.txt
@@ -0,0 +1,4 @@
+line1
+line2
+added line 1
+added line 2
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/016_pure_addition_update_chunk/input/input.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/016_pure_addition_update_chunk/input/input.txt
@@ -0,0 +1,2 @@
+line1
+line2
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/016_pure_addition_update_chunk/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/016_pure_addition_update_chunk/patch.txt
@@ -0,0 +1,6 @@
+*** Begin Patch
+*** Update File: input.txt
+@@
+added line 1
+added line 2
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/017_whitespace_padded_hunk_header/expected/foo.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/017_whitespace_padded_hunk_header/expected/foo.txt
@@ -0,0 +1 @@
+new
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/017_whitespace_padded_hunk_header/input/foo.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/017_whitespace_padded_hunk_header/input/foo.txt
@@ -0,0 +1 @@
+old
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/017_whitespace_padded_hunk_header/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/017_whitespace_padded_hunk_header/patch.txt
@@ -0,0 +1,6 @@
+*** Begin Patch
+  *** Update File: foo.txt
+@@
+-old
+new
+*** End Patch
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/018_whitespace_padded_patch_markers/expected/file.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/018_whitespace_padded_patch_markers/expected/file.txt
@@ -0,0 +1 @@
+two
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/018_whitespace_padded_patch_markers/input/file.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/018_whitespace_padded_patch_markers/input/file.txt
@@ -0,0 +1 @@
+one
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/018_whitespace_padded_patch_markers/patch.txt
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/018_whitespace_padded_patch_markers/patch.txt
@@ -0,0 +1,6 @@
+ *** Begin Patch
+*** Update File: file.txt
+@@
+-one
+two
+*** End Patch 
--- a/codex-rs/apply-patch/tests/fixtures/scenarios/README.md
+++ b/codex-rs/apply-patch/tests/fixtures/scenarios/README.md
@@ -0,0 +1,18 @@
+# Overview
+This directory is a collection of end to end tests for the apply-patch specification, meant to be easily portable to other languages or platforms.
+
+
+# Specification
+Each test case is one directory, composed of input state (input/), the patch operation (patch.txt), and the expected final state (expected/). This structure is designed to keep tests simple (i.e. test exactly one patch at a time) while still providing enough flexibility to test any given operation across files.
+
+Here's what this would look like for a simple test apply-patch test case to create a new file:
+
+```
+001_add/
+  input/
+    foo.md
+  expected/
+    foo.md
+    bar.md
+  patch.txt
+```
--- a/codex-rs/apply-patch/tests/suite/mod.rs
+++ b/codex-rs/apply-patch/tests/suite/mod.rs
@@ -1,3 +1,4 @@
 mod cli;
+mod scenarios;
 #[cfg(not(target_os = "windows"))]
 mod tool;
--- a/codex-rs/apply-patch/tests/suite/scenarios.rs
+++ b/codex-rs/apply-patch/tests/suite/scenarios.rs
@@ -0,0 +1,114 @@
+use assert_cmd::prelude::*;
+use pretty_assertions::assert_eq;
+use std::collections::BTreeMap;
+use std::fs;
+use std::path::Path;
+use std::path::PathBuf;
+use std::process::Command;
+use tempfile::tempdir;
+
+#[test]
+fn test_apply_patch_scenarios() -> anyhow::Result<()> {
+    for scenario in fs::read_dir("tests/fixtures/scenarios")? {
+        let scenario = scenario?;
+        let path = scenario.path();
+        if path.is_dir() {
+            run_apply_patch_scenario(&path)?;
+        }
+    }
+    Ok(())
+}
+
+/// Reads a scenario directory, copies the input files to a temporary directory, runs apply-patch,
+/// and asserts that the final state matches the expected state exactly.
+fn run_apply_patch_scenario(dir: &Path) -> anyhow::Result<()> {
+    let tmp = tempdir()?;
+
+    // Copy the input files to the temporary directory
+    let input_dir = dir.join("input");
+    if input_dir.is_dir() {
+        copy_dir_recursive(&input_dir, tmp.path())?;
+    }
+
+    // Read the patch.txt file
+    let patch = fs::read_to_string(dir.join("patch.txt"))?;
+
+    // Run apply_patch in the temporary directory. We intentionally do not assert
+    // on the exit status here; the scenarios are specified purely in terms of
+    // final filesystem state, which we compare below.
+    Command::cargo_bin("apply_patch")?
+        .arg(patch)
+        .current_dir(tmp.path())
+        .output()?;
+
+    // Assert that the final state matches the expected state exactly
+    let expected_dir = dir.join("expected");
+    let expected_snapshot = snapshot_dir(&expected_dir)?;
+    let actual_snapshot = snapshot_dir(tmp.path())?;
+
+    assert_eq!(
+        actual_snapshot,
+        expected_snapshot,
+        "Scenario {} did not match expected final state",
+        dir.display()
+    );
+
+    Ok(())
+}
+
+#[derive(Debug, Clone, PartialEq, Eq)]
+enum Entry {
+    File(Vec<u8>),
+    Dir,
+}
+
+fn snapshot_dir(root: &Path) -> anyhow::Result<BTreeMap<PathBuf, Entry>> {
+    let mut entries = BTreeMap::new();
+    if root.is_dir() {
+        snapshot_dir_recursive(root, root, &mut entries)?;
+    }
+    Ok(entries)
+}
+
+fn snapshot_dir_recursive(
+    base: &Path,
+    dir: &Path,
+    entries: &mut BTreeMap<PathBuf, Entry>,
+) -> anyhow::Result<()> {
+    for entry in fs::read_dir(dir)? {
+        let entry = entry?;
+        let path = entry.path();
+        let Some(stripped) = path.strip_prefix(base).ok() else {
+            continue;
+        };
+        let rel = stripped.to_path_buf();
+        let file_type = entry.file_type()?;
+        if file_type.is_dir() {
+            entries.insert(rel.clone(), Entry::Dir);
+            snapshot_dir_recursive(base, &path, entries)?;
+        } else if file_type.is_file() {
+            let contents = fs::read(&path)?;
+            entries.insert(rel, Entry::File(contents));
+        }
+    }
+    Ok(())
+}
+
+fn copy_dir_recursive(src: &Path, dst: &Path) -> anyhow::Result<()> {
+    for entry in fs::read_dir(src)? {
+        let entry = entry?;
+        let path = entry.path();
+        let file_type = entry.file_type()?;
+        let dest_path = dst.join(entry.file_name());
+        if file_type.is_dir() {
+            fs::create_dir_all(&dest_path)?;
+            copy_dir_recursive(&path, &dest_path)?;
+        } else if file_type.is_file() {
+            if let Some(parent) = dest_path.parent() {
+                fs::create_dir_all(parent)?;
+            }
+            fs::copy(&path, &dest_path)?;
+        }
+    }
+    Ok(())
+}
--- a/codex-rs/backend-client/src/client.rs
+++ b/codex-rs/backend-client/src/client.rs
@@ -7,6 +7,7 @@ use crate::types::TurnAttemptsSiblingTurnsResponse;
 use anyhow::Result;
 use codex_core::auth::CodexAuth;
 use codex_core::default_client::get_codex_user_agent;
+use codex_protocol::account::PlanType as AccountPlanType;
 use codex_protocol::protocol::CreditsSnapshot;
 use codex_protocol::protocol::RateLimitSnapshot;
 use codex_protocol::protocol::RateLimitWindow;
@@ -291,6 +292,7 @@ impl Client {
            primary,
            secondary,
            credits: Self::map_credits(payload.credits),
+            plan_type: Some(Self::map_plan_type(payload.plan_type)),
        }
    }

@@ -325,6 +327,23 @@ impl Client {
        })
    }

+    fn map_plan_type(plan_type: crate::types::PlanType) -> AccountPlanType {
+        match plan_type {
+            crate::types::PlanType::Free => AccountPlanType::Free,
+            crate::types::PlanType::Plus => AccountPlanType::Plus,
+            crate::types::PlanType::Pro => AccountPlanType::Pro,
+            crate::types::PlanType::Team => AccountPlanType::Team,
+            crate::types::PlanType::Business => AccountPlanType::Business,
+            crate::types::PlanType::Enterprise => AccountPlanType::Enterprise,
+            crate::types::PlanType::Edu | crate::types::PlanType::Education => AccountPlanType::Edu,
+            crate::types::PlanType::Guest
+            | crate::types::PlanType::Go
+            | crate::types::PlanType::FreeWorkspace
+            | crate::types::PlanType::Quorum
+            | crate::types::PlanType::K12 => AccountPlanType::Unknown,
+        }
+    }
+
    fn window_minutes_from_seconds(seconds: i32) -> Option<i64> {
        if seconds <= 0 {
            return None;
--- a/codex-rs/cli/Cargo.toml
+++ b/codex-rs/cli/Cargo.toml
@@ -36,6 +36,7 @@ codex-responses-api-proxy = { workspace = true }
 codex-rmcp-client = { workspace = true }
 codex-stdio-to-uds = { workspace = true }
 codex-tui = { workspace = true }
+codex-tui2 = { workspace = true }
 ctor = { workspace = true }
 libc = { workspace = true }
 owo-colors = { workspace = true }
--- a/codex-rs/cli/src/main.rs
+++ b/codex-rs/cli/src/main.rs
@@ -18,11 +18,14 @@ use codex_cli::login::run_logout;
 use codex_cloud_tasks::Cli as CloudTasksCli;
 use codex_common::CliConfigOverrides;
 use codex_exec::Cli as ExecCli;
+use codex_exec::Command as ExecCommand;
+use codex_exec::ReviewArgs;
 use codex_execpolicy::ExecPolicyCheckCommand;
 use codex_responses_api_proxy::Args as ResponsesApiProxyArgs;
 use codex_tui::AppExitInfo;
 use codex_tui::Cli as TuiCli;
 use codex_tui::update_action::UpdateAction;
+use codex_tui2 as tui2;
 use owo_colors::OwoColorize;
 use std::path::PathBuf;
 use supports_color::Stream;
@@ -35,6 +38,11 @@ use crate::mcp_cmd::McpCli;

 use codex_core::config::Config;
 use codex_core::config::ConfigOverrides;
+use codex_core::config::find_codex_home;
+use codex_core::config::load_config_as_toml_with_cli_overrides;
+use codex_core::features::Feature;
+use codex_core::features::FeatureOverrides;
+use codex_core::features::Features;
 use codex_core::features::is_known_feature_key;

 /// Codex CLI
@@ -72,6 +80,9 @@ enum Subcommand {
    #[clap(visible_alias = "e")]
    Exec(ExecCli),

+    /// Run a code review non-interactively.
+    Review(ReviewArgs),
+
    /// Manage login.
    Login(LoginCommand),

@@ -439,7 +450,7 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
                &mut interactive.config_overrides,
                root_config_overrides.clone(),
            );
-            let exit_info = codex_tui::run_main(interactive, codex_linux_sandbox_exe).await?;
+            let exit_info = run_interactive_tui(interactive, codex_linux_sandbox_exe).await?;
            handle_app_exit(exit_info)?;
        }
        Some(Subcommand::Exec(mut exec_cli)) => {
@@ -449,6 +460,15 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
            );
            codex_exec::run_main(exec_cli, codex_linux_sandbox_exe).await?;
        }
+        Some(Subcommand::Review(review_args)) => {
+            let mut exec_cli = ExecCli::try_parse_from(["codex", "exec"])?;
+            exec_cli.command = Some(ExecCommand::Review(review_args));
+            prepend_config_flags(
+                &mut exec_cli.config_overrides,
+                root_config_overrides.clone(),
+            );
+            codex_exec::run_main(exec_cli, codex_linux_sandbox_exe).await?;
+        }
        Some(Subcommand::McpServer) => {
            codex_mcp_server::run_main(codex_linux_sandbox_exe, root_config_overrides).await?;
        }
@@ -485,7 +505,7 @@ async fn cli_main(codex_linux_sandbox_exe: Option<PathBuf>) -> anyhow::Result<()
                all,
                config_overrides,
            );
-            let exit_info = codex_tui::run_main(interactive, codex_linux_sandbox_exe).await?;
+            let exit_info = run_interactive_tui(interactive, codex_linux_sandbox_exe).await?;
            handle_app_exit(exit_info)?;
        }
        Some(Subcommand::Login(mut login_cli)) => {
@@ -636,6 +656,40 @@ fn prepend_config_flags(
        .splice(0..0, cli_config_overrides.raw_overrides);
 }

+/// Run the interactive Codex TUI, dispatching to either the legacy implementation or the
+/// experimental TUI v2 shim based on feature flags resolved from config.
+async fn run_interactive_tui(
+    interactive: TuiCli,
+    codex_linux_sandbox_exe: Option<PathBuf>,
+) -> std::io::Result<AppExitInfo> {
+    if is_tui2_enabled(&interactive).await? {
+        let result = tui2::run_main(interactive.into(), codex_linux_sandbox_exe).await?;
+        Ok(result.into())
+    } else {
+        codex_tui::run_main(interactive, codex_linux_sandbox_exe).await
+    }
+}
+
+/// Returns `Ok(true)` when the resolved configuration enables the `tui2` feature flag.
+///
+/// This performs a lightweight config load (honoring the same precedence as the lower-level TUI
+/// bootstrap: `$CODEX_HOME`, config.toml, profile, and CLI `-c` overrides) solely to decide which
+/// TUI frontend to launch. The full configuration is still loaded later by the interactive TUI.
+async fn is_tui2_enabled(cli: &TuiCli) -> std::io::Result<bool> {
+    let raw_overrides = cli.config_overrides.raw_overrides.clone();
+    let overrides_cli = codex_common::CliConfigOverrides { raw_overrides };
+    let cli_kv_overrides = overrides_cli
+        .parse_overrides()
+        .map_err(|e| std::io::Error::new(std::io::ErrorKind::InvalidInput, e))?;
+
+    let codex_home = find_codex_home()?;
+    let config_toml = load_config_as_toml_with_cli_overrides(&codex_home, cli_kv_overrides).await?;
+    let config_profile = config_toml.get_config_profile(cli.config_profile.clone())?;
+    let overrides = FeatureOverrides::default();
+    let features = Features::from_config(&config_toml, &config_profile, overrides);
+    Ok(features.enabled(Feature::Tui2))
+}
+
 /// Build the final `TuiCli` for a `codex resume` invocation.
 fn finalize_resume_interactive(
    mut interactive: TuiCli,
--- a/codex-rs/cli/tests/execpolicy.rs
+++ b/codex-rs/cli/tests/execpolicy.rs
@@ -40,17 +40,15 @@ prefix_rule(
    assert_eq!(
        result,
        json!({
-            "match": {
-                "decision": "forbidden",
-                "matchedRules": [
-                    {
-                        "prefixRuleMatch": {
-                            "matchedPrefix": ["git", "push"],
-                            "decision": "forbidden"
-                        }
+            "decision": "forbidden",
+            "matchedRules": [
+                {
+                    "prefixRuleMatch": {
+                        "matchedPrefix": ["git", "push"],
+                        "decision": "forbidden"
                    }
-                ]
-            }
+                }
+            ]
        })
    );

--- a/codex-rs/client.md
+++ b/codex-rs/client.md
@@ -1,206 +0,0 @@
-# Client Extraction Plan
-
-## Goals
- Split the HTTP transport/client code out of `codex-core` into a reusable crate that is agnostic of Codex/OpenAI business logic and API schemas.
- Create a separate API library crate that houses typed requests/responses for well-known APIs (Responses, Chat Completions, Compact) and plugs into the transport crate via minimal traits.
- Preserve current behaviour (auth headers, retries, SSE handling, rate-limit parsing, compaction, fixtures) while making the APIs symmetric and avoiding code duplication.
- Keep existing consumers (`codex-core`, tests, and tools) stable by providing a small compatibility layer during the transition.
-
-## Snapshot of Today
- `core/src/client.rs (ModelClient)` owns config/auth/session state, chooses wire API, builds payloads, drives retries, parses SSE, compaction, and rate-limit headers.
- `core/src/chat_completions.rs` implements the Chat Completions call + SSE parser + aggregation helper.
- `core/src/client_common.rs` holds `Prompt`, tool specs, shared request structs (`ResponsesApiRequest`, `TextControls`), and `ResponseEvent`/`ResponseStream`.
- `core/src/default_client.rs` wraps `reqwest` with Codex UA/originator defaults.
- `core/src/model_provider_info.rs` models providers (base URL, headers, env keys, retry/timeout tuning) and builds `CodexRequestBuilder`s.
-    - Current retry logic is co-located with API handling; streaming SSE parsing is duplicated across Responses/Chat.
-
-## Target Crates (with interfaces)
-
- `codex-client` (generic transport)
-  - Owns the generic HTTP machinery: a `CodexHttpClient`/`CodexRequestBuilder`-style wrapper, retry/backoff hooks, streaming connector (SSE framing + idle timeout), header injection, and optional telemetry callbacks.
-  - Does **not** know about OpenAI/Codex-specific paths, headers, or error codes; it only exposes HTTP-level concepts (status, headers, bodies, connection errors).
-  - Minimal surface:
-    ```rust
-    pub trait HttpTransport {
-        fn execute(&self, req: Request) -> Result<Response, TransportError>;
-        fn stream(&self, req: Request) -> Result<ByteStream, TransportError>;
-    }
-
-    pub struct Request {
-        pub method: Method,
-        pub url: String,
-        pub headers: HeaderMap,
-        pub body: Option<serde_json::Value>,
-        pub timeout: Option<Duration>,
-    }
-    ```
-  - Generic client traits (request/response/chunk are abstract over the transport):
-    ```rust
-    #[async_trait::async_trait]
-    pub trait UnaryClient<Req, Resp> {
-        async fn run(&self, req: Req) -> Result<Resp, TransportError>;
-    }
-
-    #[async_trait::async_trait]
-    pub trait StreamClient<Req, Chunk> {
-        async fn run(&self, req: Req) -> Result<ResponseStream<Chunk>, TransportError>;
-    }
-
-    pub struct RetryPolicy {
-        pub max_attempts: u64,
-        pub base_delay: Duration,
-        pub retry_on: RetryOn, // e.g., transport errors + 429/5xx
-    }
-    ```
-    - `RetryOn` lives in `codex-client` and captures HTTP status classes and transport failures that qualify for retry.
-    - Implementations in `codex-api` plug in their own request types, parsers, and retry policies while reusing the transport’s backoff and error types.
-    - Planned runtime helper:
-      ```rust
-      pub async fn run_with_retry<T, F, Fut>(
-          policy: RetryPolicy,
-          make_req: impl Fn() -> Request,
-          op: F,
-      ) -> Result<T, TransportError>
-      where
-          F: Fn(Request) -> Fut,
-          Fut: Future<Output = Result<T, TransportError>>,
-      {
-          for attempt in 0..=policy.max_attempts {
-              let req = make_req();
-              match op(req).await {
-                  Ok(resp) => return Ok(resp),
-                  Err(err) if policy.retry_on.should_retry(&err, attempt) => {
-                      tokio::time::sleep(backoff(policy.base_delay, attempt + 1)).await;
-                  }
-                  Err(err) => return Err(err),
-              }
-          }
-          Err(TransportError::RetryLimit)
-      }
-      ```
-      - Unary clients wrap `transport.execute` with this helper and then deserialize.
-      - Stream clients wrap the **initial** `transport.stream` call with this helper. Mid-stream disconnects are surfaced as `StreamError`s; automatic resume/reconnect can be added later on top of this primitive if we introduce cursor support.
-  - Common helpers: `retry::backoff(attempt)`, `errors::{TransportError, StreamError}`.
-  - Streaming utility (SSE framing only):
-    ```rust
-    pub fn sse_stream<S>(
-        bytes: S,
-        idle_timeout: Duration,
-        tx: mpsc::Sender<Result<String, StreamError>>,
-        telemetry: Option<Box<dyn Telemetry>>,
-    )
-    where
-        S: Stream<Item = Result<Bytes, TransportError>> + Unpin + Send + 'static;
-    ```
-    - `sse_stream` is responsible for timeouts, connection-level errors, and emitting raw `data:` chunks as UTF-8 strings; parsing those strings into structured events is done in `codex-api`.
-
- `codex-api` (OpenAI/Codex API library)
-  - Owns typed models for Responses/Chat/Compact plus shared helpers (`Prompt`, tool specs, text controls, `ResponsesApiRequest`, etc.).
-  - Knows about OpenAI/Codex semantics:
-    - URL shapes (`/v1/responses`, `/v1/chat/completions`, `/responses/compact`).
-    - Provider configuration (`WireApi`, base URLs, query params, per-provider retry knobs).
-    - Rate-limit headers (`x-codex-*`) and their mapping into `RateLimitSnapshot` / `CreditsSnapshot`.
-    - Error body formats (`{ error: { type, code, message, plan_type, resets_at } }`) and how they become API errors (context window exceeded, quota/usage limit, etc.).
-    - SSE event names (`response.output_item.done`, `response.completed`, `response.failed`, etc.) and their mapping into high-level events.
-  - Provides a provider abstraction (conceptually similar to `ModelProviderInfo`):
-    ```rust
-    pub struct Provider {
-        pub name: String,
-        pub base_url: String,
-        pub wire: WireApi, // Responses | Chat
-        pub headers: HeaderMap,
-        pub retry: RetryConfig,
-        pub stream_idle_timeout: Duration,
-    }
-
-    pub trait AuthProvider {
-        /// Returns a bearer token to use for this request (if any).
-        /// Implementations are expected to be cheap and to surface already-refreshed tokens;
-        /// higher layers (`codex-core`) remain responsible for token refresh flows.
-        fn bearer_token(&self) -> Option<String>;
-
-        /// Optional ChatGPT account id header for Chat mode.
-        fn account_id(&self) -> Option<String>;
-    }
-    ```
-  - Ready-made clients built on `HttpTransport`:
-    ```rust
-    pub struct ResponsesClient<T: HttpTransport, A: AuthProvider> { /* ... */ }
-    impl<T, A> ResponsesClient<T, A> {
-        pub async fn stream(&self, prompt: &Prompt) -> ApiResult<ResponseStream<ApiEvent>>;
-        pub async fn compact(&self, prompt: &Prompt) -> ApiResult<Vec<ResponseItem>>;
-    }
-
-    pub struct ChatClient<T: HttpTransport, A: AuthProvider> { /* ... */ }
-    impl<T, A> ChatClient<T, A> {
-        pub async fn stream(&self, prompt: &Prompt) -> ApiResult<ResponseStream<ApiEvent>>;
-    }
-
-    pub struct CompactClient<T: HttpTransport, A: AuthProvider> { /* ... */ }
-    impl<T, A> CompactClient<T, A> {
-        pub async fn compact(&self, prompt: &Prompt) -> ApiResult<Vec<ResponseItem>>;
-    }
-    ```
-  - Streaming events unified across wire APIs (this can closely mirror `ResponseEvent` today, and we may type-alias one to the other during migration):
-    ```rust
-    pub enum ApiEvent {
-        Created,
-        OutputItemAdded(ResponseItem),
-        OutputItemDone(ResponseItem),
-        OutputTextDelta(String),
-        ReasoningContentDelta { delta: String, content_index: i64 },
-        ReasoningSummaryDelta { delta: String, summary_index: i64 },
-        RateLimits(RateLimitSnapshot),
-        Completed { response_id: String, token_usage: Option<TokenUsage> },
-    }
-    ```
-  - Error layering:
-    - `codex-client`: defines `TransportError` / `StreamError` (status codes, IO, timeouts).
-    - `codex-api`: defines `ApiError` that wraps `TransportError` plus API-specific errors parsed from bodies and headers.
-    - `codex-core`: maps `ApiError` into existing `CodexErr` variants so downstream callers remain unchanged.
-  - Aggregation strategies (today’s `AggregateStreamExt`) live here as adapters (`Aggregated`, `Streaming`) that transform `ResponseStream<ApiEvent>` into the higher-level views used by `codex-core`.
-
-## Implementation Steps
-
-1. **Create crates**: add `codex-client` and `codex-api` (names keep the `codex-` prefix). Stub lib files with feature flags/tests wired into the workspace; wire them into `Cargo.toml`.
-2. **Extract API-level SSE + rate limits into `codex-api`**:
-   - Move the Responses SSE parser (`process_sse`), rate-limit parsing, and related tests from `core/src/client.rs` into `codex-api`, keeping the behavior identical.
-   - Introduce `ApiEvent` (initially equivalent to `ResponseEvent`) and `ApiError`, and adjust the parser to emit those.
-   - Provide test-only helpers for fixture streams (replacement for `CODEX_RS_SSE_FIXTURE`) in `codex-api`.
-3. **Lift transport layer into `codex-client`**:
-   - Move `CodexHttpClient`/`CodexRequestBuilder`, UA/originator plumbing, and backoff helpers from `core/src/default_client.rs` into `codex-client` (or a thin wrapper on top of it).
-   - Introduce `HttpTransport`, `Request`, `RetryPolicy`, `RetryOn`, and `run_with_retry` as described above.
-   - Keep sandbox/no-proxy toggles behind injected configuration so `codex-client` stays generic and does not depend on Codex-specific env vars.
-4. **Model provider abstraction in `codex-api`**:
-   - Relocate `ModelProviderInfo` (base URL, env/header resolution, retry knobs, wire API enum) into `codex-api`, expressed in terms of `Provider` and `AuthProvider`.
-   - Ensure provider logic handles:
-     - URL building for Responses/Chat/Compact (including Azure special cases).
-     - Static and env-based headers.
-     - Per-provider retry and idle-timeout settings that map cleanly into `RetryPolicy`/`RetryOn`.
-5. **API crate wiring**:
-   - Move `Prompt`, tool specs, `ResponsesApiRequest`, `TextControls`, and `ResponseEvent/ResponseStream` into `codex-api` under modules (`common`, `responses`, `chat`, `compact`), keeping public types stable or re-exported through `codex-core` as needed.
-   - Rebuild Responses and Chat clients on top of `HttpTransport` + `StreamClient`, reusing shared retry + SSE helpers; keep aggregation adapters as reusable strategies instead of `ModelClient`-local logic.
-   - Implement Compact on top of `UnaryClient` and the unary `execute` path with JSON deserialization, sharing the same retry policy.
-   - Keep request builders symmetric: each client prepares a `Request<serde_json::Value>`, attaches headers/auth via `AuthProvider`, and plugs in its parser (streaming clients) or deserializer (unary) while sharing retry/backoff configuration derived from `Provider`.
-6. **Core integration layer**:
-   - Replace `core::ModelClient` internals with thin adapters that construct `codex-api` clients using `Config`, `AuthManager`, and `OtelEventManager`.
-   - Keep the public `ModelClient` API and `ResponseEvent`/`ResponseStream` types stable by re-exporting `codex-api` types or providing type aliases.
-   - Preserve existing auth flows (including ChatGPT token refresh) inside `codex-core` or a thin adapter, using `AuthProvider` to surface bearer tokens to `codex-api` and handling 401/refresh semantics at this layer.
-7. **Tests/migration**:
-   - Move unit tests for SSE parsing, retry/backoff decisions, and provider/header behavior into the new crates; keep integration tests in `core` using the compatibility layer.
-   - Update fixtures to be consumed via test-only adapters in `codex-api`.
-   - Run targeted `just fmt`, `just fix -p` for the touched crates, and scoped `cargo test -p codex-client`, `-p codex-api`, and existing `codex-core` suites.
-
-## Design Decisions
-
- **UA construction**
-  - `codex-client` exposes an optional UA suffix/provider hook (tiny feature) and remains unaware of the CLI; `codex-core` / the CLI compute the full UA (including `terminal::user_agent()`) and pass the suffix or builder down.
- **Config vs provider**
-  - Most configuration stays in `codex-core`. `codex-api::Provider` only contains what is strictly required for HTTP (base URLs, query params, retry/timeout knobs, wire API), while higher-level knobs (reasoning defaults, verbosity flags, etc.) remain core concerns.
- **Auth flow ownership**
-  - Auth flows (including ChatGPT token refresh) remain in `codex-core`. `AuthProvider` simply exposes already-fresh tokens/account IDs; 401 handling and refresh retries stay in the existing auth layer.
- **Error enums**
-  - `codex-client` continues to define `TransportError` / `StreamError`. `codex-api` defines an `ApiError` (deriving `thiserror::Error`) that wraps `TransportError` and API-specific failures, and `codex-core` maps `ApiError` into existing `CodexErr` variants for callers.
- **Streaming reconnection semantics**
-  - For now, mid-stream SSE failures are surfaced as errors and only the initial connection is retried via `run_with_retry`. We will revisit mid-stream reconnect/resume once the underlying APIs support cursor/idempotent event semantics.
-
--- a/codex-rs/cloud-tasks-client/src/api.rs
+++ b/codex-rs/cloud-tasks-client/src/api.rs
@@ -127,6 +127,7 @@ impl Default for TaskText {
 #[async_trait::async_trait]
 pub trait CloudBackend: Send + Sync {
    async fn list_tasks(&self, env: Option<&str>) -> Result<Vec<TaskSummary>>;
+    async fn get_task_summary(&self, id: TaskId) -> Result<TaskSummary>;
    async fn get_task_diff(&self, id: TaskId) -> Result<Option<String>>;
    /// Return assistant output messages (no diff) when available.
    async fn get_task_messages(&self, id: TaskId) -> Result<Vec<String>>;
--- a/codex-rs/cloud-tasks-client/src/http.rs
+++ b/codex-rs/cloud-tasks-client/src/http.rs
@@ -63,6 +63,10 @@ impl CloudBackend for HttpClient {
        self.tasks_api().list(env).await
    }

+    async fn get_task_summary(&self, id: TaskId) -> Result<TaskSummary> {
+        self.tasks_api().summary(id).await
+    }
+
    async fn get_task_diff(&self, id: TaskId) -> Result<Option<String>> {
        self.tasks_api().diff(id).await
    }
@@ -149,6 +153,75 @@ mod api {
            Ok(tasks)
        }

+        pub(crate) async fn summary(&self, id: TaskId) -> Result<TaskSummary> {
+            let id_str = id.0.clone();
+            let (details, body, ct) = self
+                .details_with_body(&id.0)
+                .await
+                .map_err(|e| CloudTaskError::Http(format!("get_task_details failed: {e}")))?;
+            let parsed: Value = serde_json::from_str(&body).map_err(|e| {
+                CloudTaskError::Http(format!(
+                    "Decode error for {}: {e}; content-type={ct}; body={body}",
+                    id.0
+                ))
+            })?;
+            let task_obj = parsed
+                .get("task")
+                .and_then(Value::as_object)
+                .ok_or_else(|| {
+                    CloudTaskError::Http(format!("Task metadata missing from details for {id_str}"))
+                })?;
+            let status_display = parsed
+                .get("task_status_display")
+                .or_else(|| task_obj.get("task_status_display"))
+                .and_then(Value::as_object)
+                .map(|m| {
+                    m.iter()
+                        .map(|(k, v)| (k.clone(), v.clone()))
+                        .collect::<HashMap<String, Value>>()
+                });
+            let status = map_status(status_display.as_ref());
+            let mut summary = diff_summary_from_status_display(status_display.as_ref());
+            if summary.files_changed == 0
+                && summary.lines_added == 0
+                && summary.lines_removed == 0
+                && let Some(diff) = details.unified_diff()
+            {
+                summary = diff_summary_from_diff(&diff);
+            }
+            let updated_at_raw = task_obj
+                .get("updated_at")
+                .and_then(Value::as_f64)
+                .or_else(|| task_obj.get("created_at").and_then(Value::as_f64))
+                .or_else(|| latest_turn_timestamp(status_display.as_ref()));
+            let environment_id = task_obj
+                .get("environment_id")
+                .and_then(Value::as_str)
+                .map(str::to_string);
+            let environment_label = env_label_from_status_display(status_display.as_ref());
+            let attempt_total = attempt_total_from_status_display(status_display.as_ref());
+            let title = task_obj
+                .get("title")
+                .and_then(Value::as_str)
+                .unwrap_or("<untitled>")
+                .to_string();
+            let is_review = task_obj
+                .get("is_review")
+                .and_then(Value::as_bool)
+                .unwrap_or(false);
+            Ok(TaskSummary {
+                id,
+                title,
+                status,
+                updated_at: parse_updated_at(updated_at_raw.as_ref()),
+                environment_id,
+                environment_label,
+                summary,
+                is_review,
+                attempt_total,
+            })
+        }
+
        pub(crate) async fn diff(&self, id: TaskId) -> Result<Option<String>> {
            let (details, body, ct) = self
                .details_with_body(&id.0)
@@ -679,6 +752,34 @@ mod api {
            .map(str::to_string)
    }

+    fn diff_summary_from_diff(diff: &str) -> DiffSummary {
+        let mut files_changed = 0usize;
+        let mut lines_added = 0usize;
+        let mut lines_removed = 0usize;
+        for line in diff.lines() {
+            if line.starts_with("diff --git ") {
+                files_changed += 1;
+                continue;
+            }
+            if line.starts_with("+++") || line.starts_with("---") || line.starts_with("@@") {
+                continue;
+            }
+            match line.as_bytes().first() {
+                Some(b'+') => lines_added += 1,
+                Some(b'-') => lines_removed += 1,
+                _ => {}
+            }
+        }
+        if files_changed == 0 && !diff.trim().is_empty() {
+            files_changed = 1;
+        }
+        DiffSummary {
+            files_changed,
+            lines_added,
+            lines_removed,
+        }
+    }
+
    fn diff_summary_from_status_display(v: Option<&HashMap<String, Value>>) -> DiffSummary {
        let mut out = DiffSummary::default();
        let Some(map) = v else { return out };
@@ -700,6 +801,17 @@ mod api {
        out
    }

+    fn latest_turn_timestamp(v: Option<&HashMap<String, Value>>) -> Option<f64> {
+        let map = v?;
+        let latest = map
+            .get("latest_turn_status_display")
+            .and_then(Value::as_object)?;
+        latest
+            .get("updated_at")
+            .or_else(|| latest.get("created_at"))
+            .and_then(Value::as_f64)
+    }
+
    fn attempt_total_from_status_display(v: Option<&HashMap<String, Value>>) -> Option<usize> {
        let map = v?;
        let latest = map
--- a/codex-rs/cloud-tasks-client/src/mock.rs
+++ b/codex-rs/cloud-tasks-client/src/mock.rs
@@ -1,6 +1,7 @@
 use crate::ApplyOutcome;
 use crate::AttemptStatus;
 use crate::CloudBackend;
+use crate::CloudTaskError;
 use crate::DiffSummary;
 use crate::Result;
 use crate::TaskId;
@@ -60,6 +61,14 @@ impl CloudBackend for MockClient {
        Ok(out)
    }

+    async fn get_task_summary(&self, id: TaskId) -> Result<TaskSummary> {
+        let tasks = self.list_tasks(None).await?;
+        tasks
+            .into_iter()
+            .find(|t| t.id == id)
+            .ok_or_else(|| CloudTaskError::Msg(format!("Task {} not found (mock)", id.0)))
+    }
+
    async fn get_task_diff(&self, id: TaskId) -> Result<Option<String>> {
        Ok(Some(mock_diff_for(&id)))
    }
--- a/codex-rs/cloud-tasks/Cargo.toml
+++ b/codex-rs/cloud-tasks/Cargo.toml
@@ -34,6 +34,9 @@ tokio-stream = { workspace = true }
 tracing = { workspace = true, features = ["log"] }
 tracing-subscriber = { workspace = true, features = ["env-filter"] }
 unicode-width = { workspace = true }
+owo-colors = { workspace = true, features = ["supports-colors"] }
+supports-color = { workspace = true }

 [dev-dependencies]
 async-trait = { workspace = true }
+pretty_assertions = { workspace = true }
--- a/codex-rs/cloud-tasks/src/app.rs
+++ b/codex-rs/cloud-tasks/src/app.rs
@@ -350,6 +350,7 @@ pub enum AppEvent {
 mod tests {
    use super::*;
    use chrono::Utc;
+    use codex_cloud_tasks_client::CloudTaskError;

    struct FakeBackend {
        // maps env key to titles
@@ -385,6 +386,17 @@ mod tests {
            Ok(out)
        }

+        async fn get_task_summary(
+            &self,
+            id: TaskId,
+        ) -> codex_cloud_tasks_client::Result<TaskSummary> {
+            self.list_tasks(None)
+                .await?
+                .into_iter()
+                .find(|t| t.id == id)
+                .ok_or_else(|| CloudTaskError::Msg(format!("Task {} not found", id.0)))
+        }
+
        async fn get_task_diff(
            &self,
            _id: TaskId,
--- a/Show More
+++ b/Show More