## Why
MCP tools were only placed behind `tool_search` when a feature flag was
enabled or when there were at least 100 tools. That made the model's
tool flow depend on both rollout configuration and the number of
installed tools.
The searched-tool flow is now the intended behavior. Making it
unconditional when the model and provider support it gives every
supported setup the same behavior and lets us retire the feature flag
safely.
## What changed
- Defer all effective MCP tools when `tool_search` and namespaced tools
are supported.
- Keep exposing MCP tools directly when search cannot be used, so older
or unsupported model/provider combinations still work.
- Mark `tool_search_always_defer_mcp_tools` as removed and ignore old
configured values.
- Keep plugin filtering, app-only filtering, file handling, and MCP
calls working through the searched-tool flow.
## Why many tests changed
Many tests used to act as if the model could see MCP tools in its first
request and call them immediately. That is no longer the real flow: the
model first receives `tool_search`, searches for a tool, receives the
matching MCP tool, and then calls it in the next request.
The tests therefore needed an extra search step, and checks for tool
names, descriptions, and input fields had to move from the first request
to the search result. These are not separate product changes; they make
the tests follow what the model will actually see after this change. The
plugin tests still check which tools are allowed and where they came
from, the file tests still check upload fields and behavior, and the MCP
round-trip test still checks a successful call from start to finish.
## Tests
- `just test -p codex-features`
- Focused `codex-core` tests for MCP exposure and tool planning
- `just test -p codex-core explicit_plugin_mentions`
- `just test -p codex-core stdio_server_round_trip`
- Focused `codex-core` tests for tool search, app-only tools, and MCP
file uploads
The workspace denies `clippy::expect_used` in production. Although
`clippy.toml` allows `expect` in tests, Bazel Clippy compiles
integration-test helper code in a way that does not receive that
exemption, which encouraged verbose `unwrap_or_else(... panic!(...))`
and equivalent `match`/`let else` forms.
This allows `clippy::expect_used` once at each integration-test crate
root (including aggregated suites and test-support libraries), then
replaces manual panic-based Result and Option unwraps with
`expect`/`expect_err`. Standalone `tests/*.rs` files remain their own
crate roots. Intentional assertion and unexpected-variant panics remain
unchanged, and the production `expect_used = "deny"` lint remains in
place.
The cleanup is mechanical and net-negative in line count.
## Context
This is the next step in the plugin auth-routing stack. The earlier PRs
make `PluginsManager` auth-aware and move the broad App/MCP surface
decision into that layer. This PR narrows the ChatGPT/SIWC behavior so
we only hide a plugin MCP server when it conflicts with an App
declaration of the same name.
In product terms: if a plugin exposes both an App route and MCP route
for `foo`, ChatGPT/SIWC sessions should use the App route for `foo`. If
the same plugin also exposes a separate MCP server like `foo2`, that MCP
server should remain available.
```json
// .app.json
{
"apps": {
"foo": {
"id": "connector_abc"
}
}
}
```
```json
// .mcp.json
{
"mcpServers": {
"foo": {
"url": "https://mcp.foo.com/mcp"
},
"foo2": {
"url": "https://mcp.foo2.com/mcp"
}
}
}
```
## Stack
- PR1: #27652 seed plugin manager auth at construction.
- PR2: #27459 route plugin surfaces by auth mode.
- PR3: #27607 dedupe plugin MCP servers by App declaration name.
- PR4: #27602 preserve plugin Apps in connector listings.
- PR5: #27461 skip install-time plugin MCP OAuth for matching App
routes.
## Summary
- Preserve App declaration names in loaded plugin metadata.
- Keep public effective App outputs as deduped connector IDs for
existing callers.
- For ChatGPT/SIWC, suppress only plugin MCP servers whose names match
declared App names.
## Validation
```bash
cargo fmt --all
cargo test -p codex-core-plugins plugin_auth_projection
cargo test -p codex-core-plugins effective_apps
cargo test -p codex-core-plugins read_plugin_for_config_installed_git_source_reads_from_cache_without_cloning
cargo test -p codex-core explicit_plugin_mentions_use_apps_for_chatgpt_dual_surface_plugins
cargo test -p codex-core explicit_plugin_mentions_keep_non_conflicting_mcp_for_chatgpt_auth
cargo test -p codex-app-server --test all plugin_install_filters_disallowed_apps_needing_auth
git diff --check
```
---------
Co-authored-by: Xin Lin <xl@openai.com>
## Context
Some plugins expose both Apps and MCP servers. This PR moves auth-aware
surface projection into `core-plugins::PluginsManager`, so callers get a
consistent effective plugin view. Later PRs narrow the conflict rule and
update listing/install paths.
The high level goal of this PR is to set up the plumbing to
conditionally filter App/MCP in the plugin manager layer. We start by
removing MCP servers when using SIWC/Codex-backend auth, and removing
Apps when using API-key-style auth.
This PR is now stacked on #27652, which contains only the constructor
plumbing for seeding `PluginsManager` with the current auth mode.
## Stack
- PR1: #27652 seed plugin manager auth at construction.
- PR2: #27459 route plugin surfaces by auth mode.
- PR3: #27607 dedupe plugin MCP servers by App declaration name.
- PR4: #27602 preserve plugin Apps in connector listings.
- PR5: #27461 skip install-time plugin MCP OAuth for matching App
routes.
## Summary
- API-key/non-ChatGPT routes hide plugin Apps and keep plugin MCPs.
- ChatGPT/SIWC with Apps enabled keeps plugin Apps and suppresses MCPs
for dual-surface plugins.
- MCP-only plugins stay available for ChatGPT/SIWC sessions.
- Cached plugin load outcomes are re-projected when auth mode changes.
## Validation
```bash
cargo test -p codex-core-plugins plugin_auth_projection
cargo test -p codex-core list_tool_suggest_discoverable_plugins
git diff --check
```
## Summary
Removes Plugin descriptions from the dev message, since descriptions of
skills and MCPs cover the capabilities offered by the plugin.
## Testing
- [x] Updates unit tests
## Why
Thread cwd and environment selections are a single logical setting in
core: updating one without the other can silently desynchronize the
next-turn execution context. This change makes that relationship
explicit in the internal thread settings flow while preserving the
existing app-server public API shape.
## What changed
- Moved the cwd/environment pair through internal
`ThreadSettingsOverrides.environment_settings` instead of a top-level
internal `cwd` field.
- Kept `thread/settings/update` public params unchanged, with app-server
translating top-level `cwd` into the paired internal settings shape.
- Moved `Op::UserInput` environment overrides into thread settings so
user turns and settings updates use the same core path.
- Updated core, app-server, MCP, memories, sample, and test callsites to
construct the paired settings shape.
## Verification
- `git diff --check`
- Local test run starting after PR creation.
## Summary
Keep the full `TestCodex` harness alive in plugin integration tests
instead of returning only the `CodexThread`.
## Why
The helper was moving a temporary `codex_home` into `TestCodex`, then
immediately dropping the harness and returning only the thread. For
plugin MCP tests, the MCP server cwd is inside that temporary home. If
the temp directory is removed while MCP startup is still racing, the
server launch can fail with `No such file or directory`.
Keeping the harness in scope keeps the temp home alive for the test
duration and removes the lifetime race behind the recent
`explicit_plugin_mentions_inject_plugin_guidance` flake.
## Validation
- `just fmt`
- `just test -p codex-core
explicit_plugin_mentions_inject_plugin_guidance`
Ensures MCP-backed `codex-core` integration tests exercise initialized
servers instead of racing server startup.
I've been idly investigating a few flakes and the failure modes are much
more confusing when a tool call fails because of a failed server start
than when the failed server start causes the test to fail directly.
## Summary
Adds experimental `additionalContext` support to `turn/start` and
`turn/steer` so clients can provide ephemeral external context, such as
browser or automation state, without turning that plumbing into a
visible user prompt or triggering user-prompt lifecycle behavior.
## API Shape
The parameter shape is:
```ts
additionalContext?: Record<string, {
value: string
kind: "untrusted" | "application"
}> | null
```
Example:
```json
{
"additionalContext": {
"browser_info": {
"value": "Active tab is CI failures.",
"kind": "untrusted"
},
"automation_info": {
"value": "CI rerun is in progress.",
"kind": "application"
}
}
}
```
The keys are opaque and caller-defined.
## Context Injection
When provided, accepted entries are inserted into model context as
hidden contextual message items, not as visible thread user-message
items.
`kind: "untrusted"` entries are inserted with role `user`:
```text
<external_${key}>${value}</external_${key}>
```
`kind: "application"` entries are inserted with role `developer`:
```text
<${key}>${value}</${key}>
```
Values are not escaped. Each value is truncated to 1k approximate tokens
before wrapping.
For `turn/start`, accepted additional context is inserted before normal
user input. For `turn/steer`, additional context is merged only when the
steer includes non-empty user input; context-only steers still reject as
empty input.
## Dedupe Strategy
`AdditionalContextStore` lives on session state and stores the latest
complete additional-context map.
Each `turn/start` or non-empty `turn/steer` treats its
`additionalContext` as the current complete set of values. Entries are
injected only when the key is new or the exact entry for that key
changed, including `value` or `kind`. After merging, the store is
replaced with the provided map, so omitted keys are removed from the
retained set and can be injected again later if reintroduced.
Omitting `additionalContext`, passing `null`, or passing an empty object
resets the store to empty and injects nothing.
## What Changed
- Threads experimental v2 `additionalContext` through app-server into
core turn start and steer handling.
- Adds separate contextual fragment types for untrusted user-role
context and application developer-role context.
- Uses pending response input items so additional context can be
combined with normal user input without treating it as prompt text.
- Adds integration coverage for start/steer flow, role routing,
dedupe/reset behavior, deletion/re-add behavior, hook-blocked input
behavior, empty context-only steer rejection, external-fragment marker
matching, and truncation.
## Why
The `non_prefixed_mcp_tool_names` feature should be applied where MCP
tools become model-visible, not by remapping names later in core.
Keeping the decision in `McpConnectionManager` construction makes
`ToolInfo` the single shaped view that spec building, deferred tool
search, routing, and unavailable-tool placeholders can consume directly.
This also preserves the existing external behavior while the feature is
off, and keeps the feature-on behavior for code mode and hooks explicit
at the manager boundary.
## What Changed
- Add `McpToolNameMode` to `codex-mcp` and flow it through `McpConfig`
into `McpConnectionManager::new`.
- Normalize MCP `ToolInfo` names in the manager using either
legacy-prefixed namespaces or non-prefixed namespaces; the legacy path
adds `mcp__` without restoring the old trailing namespace suffix.
- Remove the core-side MCP name remapping path so specs, tool search,
session resolution, and unavailable-tool placeholder construction use
the manager-provided `ToolName` values directly.
- Keep code mode flattening on the `__` namespace separator.
- Preserve hook compatibility by giving non-prefixed MCP hook names
legacy `mcp__...` matcher aliases.
- Add/adjust integration and unit coverage for non-prefixed code-mode
behavior, hook matching with the feature on and off, and manager-level
legacy prefixing.
## Testing
- `cargo test -p codex-mcp --lib`
- `cargo test -p codex-core --lib tools::spec::tests -- --nocapture`
- `cargo test -p codex-core --lib mcp_tools -- --nocapture`
- `cargo test -p codex-core --lib mcp_tool_exposure -- --nocapture`
- `cargo test -p codex-core --test all mcp_tool -- --nocapture`
- `cargo test -p codex-core --test all search_tool -- --nocapture`
- `cargo test -p codex-core --test all hooks_mcp -- --nocapture`
- `cargo test -p codex-core --test all
code_mode_uses_non_prefixed_mcp_tool_names_when_feature_enabled --
--nocapture`
- `cargo test -p codex-tools`
- `cargo test -p codex-features`
**Stack position:** [1 of 7]
## Summary
The first three PRs in this stack are a cleanup pass before the actual
thread settings API work.
Today, core has several overlapping "user input" ops: `UserInput`,
`UserInputWithTurnContext`, and `UserTurn`. They differ mostly in how
much next-turn state they carry, which makes the later queued thread
settings update harder to reason about and review.
This PR starts that cleanup by adding the shared
`ThreadSettingsOverrides` payload and allowing `Op::UserInput` to carry
it. Existing variants remain in place here, so this layer is mostly a
behavior-preserving API shape change plus mechanical constructor
updates.
## End State After PR3
By the end of PR3, `Op::UserInput` is the only "user input" core op. It
can carry optional thread settings overrides for callers that need to
update stored defaults with a turn, while callers without updates use
empty settings. `Op::UserInputWithTurnContext` and `Op::UserTurn` are
deleted.
## End State After PR5
By the end of PR5, core will have only two ops for this area:
- `Op::UserInput` for user-input-bearing submissions.
- `Op::ThreadSettings` for settings-only updates.
## Stack
1. [1 of 7] [Add thread settings to
UserInput](https://github.com/openai/codex/pull/23080) (this PR)
2. [2 of 7] [Remove
UserInputWithTurnContext](https://github.com/openai/codex/pull/23081)
3. [3 of 7] [Remove
UserTurn](https://github.com/openai/codex/pull/23075)
4. [4 of 7] [Placeholder for OverrideTurnContext
cleanup](https://github.com/openai/codex/pull/23087)
5. [5 of 7] [Replace OverrideTurnContext with
ThreadSettings](https://github.com/openai/codex/pull/22508)
6. [6 of 7] [Add app-server thread settings
API](https://github.com/openai/codex/pull/22509)
7. [7 of 7] [Sync TUI thread
settings](https://github.com/openai/codex/pull/22510)
## Why
The core `Op::ListMcpTools` request path is no longer needed. Keeping it
around left a dead request/response surface alongside the app-server MCP
inventory APIs that own current server status listing.
## What Changed
- Removed `Op::ListMcpTools`, `EventMsg::McpListToolsResponse`, and the
core handler that built the MCP snapshot response.
- Removed the now-unused `codex-mcp` snapshot wrapper/export and passive
event handling arms in rollout and MCP-server consumers.
- Updated tests that used the old op as a synchronization hook to wait
on existing startup/skills events, and deleted the plugin test that only
exercised the removed listing op.
## Validation
- `cargo test -p codex-protocol`
- `cargo test -p codex-mcp`
- `cargo test -p codex-rollout -p codex-rollout-trace -p
codex-mcp-server`
- `cargo test -p codex-core --test all
pending_input::queued_inter_agent_mail`
- `cargo test -p codex-core --test all
rmcp_client::stdio_mcp_tool_call_includes_sandbox_state_meta`
- `cargo test -p codex-core --test all
rmcp_client::stdio_image_responses`
- `just fix -p codex-core -p codex-protocol -p codex-mcp -p
codex-rollout -p codex-rollout-trace -p codex-mcp-server`
## Why
Recent `main` CI had repeated flakes in the plugin fixture tests:
- `codex-core::all
suite::plugins::explicit_plugin_mentions_inject_plugin_guidance` failed
in runs
[24909500958](https://github.com/openai/codex/actions/runs/24909500958),
[24908076251](https://github.com/openai/codex/actions/runs/24908076251),
[24906197645](https://github.com/openai/codex/actions/runs/24906197645),
and
[24898949647](https://github.com/openai/codex/actions/runs/24898949647).
- `codex-core::all suite::plugins::plugin_mcp_tools_are_listed` failed
in runs
[24909500958](https://github.com/openai/codex/actions/runs/24909500958),
[24908076251](https://github.com/openai/codex/actions/runs/24908076251),
and
[24898949647](https://github.com/openai/codex/actions/runs/24898949647).
The failures were in the same plugin/MCP fixture family: assertions
expected sample plugin guidance or tool inventory, but the test could
observe the session before the sample MCP server had finished startup.
## Root Cause
`explicit_plugin_mentions_inject_plugin_guidance` submitted the user
turn immediately after constructing the session. MCP startup is
asynchronous, so on a slower or busier CI runner the prompt could be
built before the sample plugin MCP server had reported its tools. That
made the test depend on scheduler timing rather than the fixture being
ready.
`plugin_mcp_tools_are_listed` already needed the same readiness
condition, but its wait logic was local to that test.
## What Changed
- Added a shared `wait_for_sample_mcp_ready` helper for the plugin
fixture tests.
- Wait for `McpStartupComplete` before submitting the explicit plugin
mention turn.
- Reuse the same readiness helper in the MCP tool-listing test.
## Why This Should Be Reliable
The tests now wait for the explicit readiness signal from the sample MCP
server before asserting guidance or tools derived from that server. This
removes the startup race while still exercising the real fixture path,
so the assertions should only run after the plugin inventory is
deterministic.
## Verification
- `cargo test -p codex-core --test all plugins::`
- GitHub CI for this PR is passing.
## Summary
The plugin MCP tool-listing test could hide MCP startup failures by
polling `ListMcpTools` until its own 30s deadline. If the plugin MCP
server startup had already failed or timed out, the session-owned MCP
manager would keep returning an empty tool list, so CI only reported
`discovered tools: []` instead of the startup state that mattered.
This makes the test synchronize on `McpStartupComplete` for the sample
plugin MCP server before asserting listed tools, and gives the
Bazel-launched test server a larger startup window.
## Notes
Confidence is about 80%. The source path strongly supports the RCA: a
failed MCP startup is represented as an empty tool list through
`ListMcpTools`, so the old polling contract could not distinguish "not
ready yet" from "startup already failed." I could not retrieve the CI
execution-log artifact to confirm the exact hidden startup error, but
the observed Ubuntu Bazel failure matches this path: repeated
`ListMcpTools` responses with no tools until the test-local timeout
fired.
I think this is the right solution because it keeps plugin behavior
unchanged and fixes only the test contract. Future startup failures
should now report the `McpStartupComplete` failure/cancellation instead
of timing out on an empty tool snapshot.
This test was introduced in https://github.com/openai/codex/pull/12864.
## Summary
- add experimental turn/start.environments params for per-turn
environment id + cwd selections
- pass selections through core protocol ops and resolve them with
EnvironmentManager before TurnContext creation
- treat omitted selections as default behavior, empty selections as no
environment, and non-empty selections as first environment/cwd as the
turn primary
## Testing
- ran `just fmt`
- ran `just write-app-server-schema`
- not run: unit tests for this stacked PR
---------
Co-authored-by: Codex <noreply@openai.com>
- Replace the active models-manager catalog with the deleted core
catalog contents.
- Replace stale hardcoded test model slugs with current bundled model
slugs.
- Keep this as a stacked change on top of the cleanup PR.
stacked on #17402.
MCP tools returned by `tool_search` (deferred tools) get registered in
our `ToolRegistry` with a different format than directly available
tools. this leads to two different ways of accessing MCP tools from our
tool catalog, only one of which works for each. fix this by registering
all MCP tools with the namespace format, since this info is already
available.
also, direct MCP tools are registered to responsesapi without a
namespace, while deferred MCP tools have a namespace. this means we can
receive MCP `FunctionCall`s in both formats from namespaces. fix this by
always registering MCP tools with namespace, regardless of deferral
status.
make code mode track `ToolName` provenance of tools so it can map the
literal JS function name string to the correct `ToolName` for
invocation, rather than supporting both in core.
this lets us unify to a single canonical `ToolName` representation for
each MCP tool and force everywhere to use that one, without supporting
fallbacks.
## Summary
App-server v2 already receives turn-scoped `clientMetadata`, but the
Rust app-server was dropping it before the outbound Responses request.
This change keeps the fix lightweight by threading that metadata through
the existing turn-metadata path rather than inventing a new transport.
## What we're trying to do and why
We want turn-scoped metadata from the app-server protocol layer,
especially fields like Hermes/GAAS run IDs, to survive all the way to
the actual Responses API request so it is visible in downstream
websocket request logging and analytics.
The specific bug was:
- app-server protocol uses camelCase `clientMetadata`
- Responses transport already has an existing turn metadata carrier:
`x-codex-turn-metadata`
- websocket transport already rewrites that header into
`request.request_body.client_metadata["x-codex-turn-metadata"]`
- but the Rust app-server never parsed or stored `clientMetadata`, so
nothing from the app-server request was making it into that existing
path
This PR fixes that without adding a new header or a second metadata
channel.
## How we did it
### Protocol surface
- Add optional `clientMetadata` to v2 `TurnStartParams` and
`TurnSteerParams`
- Regenerate the JSON schema / TypeScript fixtures
- Update app-server docs to describe the field and its behavior
### Runtime plumbing
- Add a dedicated core op for app-server user input carrying turn-scoped
metadata: `Op::UserInputWithClientMetadata`
- Wire `turn/start` and `turn/steer` through that op / signature path
instead of dropping the metadata at the message-processor boundary
- Store the metadata in `TurnMetadataState`
### Transport behavior
- Reuse the existing serialized `x-codex-turn-metadata` payload
- Merge the new app-server `clientMetadata` into that JSON additively
- Do **not** replace built-in reserved fields already present in the
turn metadata payload
- Keep websocket behavior unchanged at the outer shape level: it still
sends only `client_metadata["x-codex-turn-metadata"]`, but that JSON
string now contains the merged fields
- Keep HTTP fallback behavior unchanged except that the existing
`x-codex-turn-metadata` header now includes the merged fields too
### Request shape before / after
Before, a websocket `response.create` looked like:
```json
{
"type": "response.create",
"client_metadata": {
"x-codex-turn-metadata": "{\"session_id\":\"...\",\"turn_id\":\"...\"}"
}
}
```
Even if the app-server caller supplied `clientMetadata`, it was not
represented there.
After, the same request shape is preserved, but the serialized payload
now includes the new turn-scoped fields:
```json
{
"type": "response.create",
"client_metadata": {
"x-codex-turn-metadata": "{\"session_id\":\"...\",\"turn_id\":\"...\",\"fiber_run_id\":\"fiber-start-123\",\"origin\":\"gaas\"}"
}
}
```
## Validation
### Targeted tests added / updated
- protocol round-trip coverage for `clientMetadata` on `turn/start` and
`turn/steer`
- protocol round-trip coverage for `Op::UserInputWithClientMetadata`
- `TurnMetadataState` merge test proving client metadata is added
without overwriting reserved built-in fields
- websocket request-shape test proving outbound `response.create`
contains merged metadata inside
`client_metadata["x-codex-turn-metadata"]`
- app-server integration tests proving:
- `turn/start` forwards `clientMetadata` into the outbound Responses
request path
- websocket warmup + real turn request both behave correctly
- `turn/steer` updates the follow-up request metadata
### Commands run
- `just write-app-server-schema`
- `cargo test -p codex-app-server-protocol`
- `cargo test -p codex-protocol`
- `cargo test -p codex-core
turn_metadata_state_merges_client_metadata_without_replacing_reserved_fields
--lib`
- `cargo test -p codex-core --test all
responses_websocket_preserves_custom_turn_metadata_fields`
- `cargo test -p codex-app-server --test all client_metadata`
- `cargo test -p codex-app-server --test all
turn_start_forwards_client_metadata_to_responses_websocket_request_body_v2
-- --nocapture`
- `just fmt`
- `just fix -p codex-core -p codex-protocol -p codex-app-server-protocol
-p codex-app-server`
- `just fix -p codex-exec -p codex-tui-app-server`
- `just argument-comment-lint`
### Full suite note
`cargo test` in `codex-rs` still fails in:
-
`suite::v2::turn_interrupt::turn_interrupt_resolves_pending_command_approval_request`
I verified that same failure on a clean detached `HEAD` worktree with an
isolated `CARGO_TARGET_DIR`, so it is not caused by this patch.
## Why
`codex-core` was re-exporting APIs owned by sibling `codex-*` crates,
which made downstream crates depend on `codex-core` as a proxy module
instead of the actual owner crate.
Removing those forwards makes crate boundaries explicit and lets leaf
crates drop unnecessary `codex-core` dependencies. In this PR, this
reduces the dependency on `codex-core` to `codex-login` in the following
files:
```
codex-rs/backend-client/Cargo.toml
codex-rs/mcp-server/tests/common/Cargo.toml
```
## What
- Remove `codex-rs/core/src/lib.rs` re-exports for symbols owned by
`codex-login`, `codex-mcp`, `codex-rollout`, `codex-analytics`,
`codex-protocol`, `codex-shell-command`, `codex-sandboxing`,
`codex-tools`, and `codex-utils-path`.
- Delete the `default_client` forwarding shim in `codex-rs/core`.
- Update in-crate and downstream callsites to import directly from the
owning `codex-*` crate.
- Add direct Cargo dependencies where callsites now target the owner
crate, and remove `codex-core` from `codex-rs/backend-client`.
- Split the feature system into a new `codex-features` crate.
- Cut `codex-core` and workspace consumers over to the new config and
warning APIs.
Co-authored-by: Ahmed Ibrahim <219906144+aibrahim-oai@users.noreply.github.com>
Co-authored-by: Codex <noreply@openai.com>
Move the general `Apps`, `Skills` and `Plugins` instructions blocks out
of `user_instructions` and into the developer message, with new `Apps ->
Skills -> Plugins` order for better clarity.
Also wrap those sections in stable XML-style instruction tags (like
other sections) and update prompt-layout tests/snapshots. This makes the
tests less brittle in snapshot output (we can parse the sections), and
it consolidates the capability instructions in one place.
#### Tests
Updated snapshots, added tests.
`<AGENTS_MD>` disappearing in snapshots is expected: before this change,
the wrapped user-instructions message was kept alive by `Skills`
content. Now that `Skills` and `Plugins` are in the developer message,
that wrapper only appears when there is real
project-doc/user-instructions content.
---------
Co-authored-by: Charley Cunningham <ccunningham@openai.com>
- clarify app mentions are in user messages
- clarify what it means for tools to be provided via `codex_apps` MCP
- add plugin descriptions (with basic sanitization) to top-level `##
Plugins` section alongside the corresponding plugin names
- explain that skills from plugins are prefixed with `plugin_name:` in
top-level `##Plugins` section
changes to more logically organize `Apps`, `Skills`, and `Plugins`
instructions will be in a separate PR, as that shuffles dev + user
instructions in ways that change tests broadly.
### Tests
confirmed in local rollout, some new tests.
## Why
to support a new bring your own search tool in Responses
API(https://developers.openai.com/api/docs/guides/tools-tool-search#client-executed-tool-search)
we migrating our bm25 search tool to use official way to execute search
on client and communicate additional tools to the model.
## What
- replace the legacy `search_tool_bm25` flow with client-executed
`tool_search`
- add protocol, SSE, history, and normalization support for
`tool_search_call` and `tool_search_output`
- return namespaced Codex Apps search results and wire namespaced
follow-up tool calls back into MCP dispatch
#### What
Add structured `@plugin` parsing and TUI support for plugin mentions.
- Core: switch from plain-text `@display_name` parsing to structured
`plugin://...` mentions via `UserInput::Mention` and
`[$...](plugin://...)` links in text, same pattern as apps/skills.
- TUI: add plugin mention popup, autocomplete, and chips when typing
`$`. Load plugin capability summaries and feed them into the composer;
plugin mentions appear alongside skills and apps.
- Generalize mention parsing to a sigil parameter, still defaults to `$`
<img width="797" height="119" alt="image"
src="https://github.com/user-attachments/assets/f0fe2658-d908-4927-9139-73f850805ceb"
/>
Builds on #13510. Currently clients have to build their own `id` via
`plugin@marketplace` and filter plugins to show by `enabled`, but we
will add `id` and `available` as fields returned from `plugin/list`
soon.
####Tests
Added tests, verified locally.
## Note-- added plugin mentions via @, but that conflicts with file
mentions
depends and builds upon #13433.
- introduces explicit `@plugin` mentions. this injects the plugin's mcp
servers, app names, and skill name format into turn context as a dev
message.
- we do not yet have UI for these mentions, so we currently parse raw
text (as opposed to skills and apps which have UI chips, autocomplete,
etc.) this depends on a `plugins/list` app-server endpoint we can feed
the UI with, which is upcoming
- also annotate mcp and app tool descriptions with the plugin(s) they
come from. this gives the model a first class way of understanding what
tools come from which plugins, which will help implicit invocation.
### Tests
Added and updated tests, unit and integration. Also confirmed locally a
raw `@plugin` injects the dev message, and the model knows about its
apps, mcps, and skills.
### first half of changes, followed by #13510
Track plugin capabilities as derived summaries on `PluginLoadOutcome`
for enabled plugins with at least one skill/app/mcp.
Also add `Plugins` section to `user_instructions` injected on session
start. These introduce the plugins concept and list enabled plugins, but
do NOT currently include paths to enabled plugins or details on what
apps/mcps the plugins contain (current plan is to inject this on
@-mention). that can be adjusted in a follow up and based on evals.
### tests
Added/updated tests, confirmed locally that new `Plugins` section +
currently enabled plugins show up in `user_instructions`.
## Why
Enterprises can already constrain approvals, sandboxing, and web search
through `requirements.toml` and MDM, but feature flags were still only
configurable as managed defaults. That meant an enterprise could suggest
feature values, but it could not actually pin them.
This change closes that gap and makes enterprise feature requirements
behave like the other constrained settings. The effective feature set
now stays consistent with enterprise requirements during config load,
when config writes are validated, and when runtime code mutates feature
flags later in the session.
It also tightens the runtime API for managed features. `ManagedFeatures`
now follows the same constraint-oriented shape as `Constrained<T>`
instead of exposing panic-prone mutation helpers, and production code
can no longer construct it through an unconstrained `From<Features>`
path.
The PR also hardens the `compact_resume_fork` integration coverage on
Windows. After the feature-management changes,
`compact_resume_after_second_compaction_preserves_history` was
overflowing the libtest/Tokio thread stacks on Windows, so the test now
uses an explicit larger-stack harness as a pragmatic mitigation. That
may not be the ideal root-cause fix, and it merits a parallel
investigation into whether part of the async future chain should be
boxed to reduce stack pressure instead.
## What Changed
Enterprises can now pin feature values in `requirements.toml` with the
requirements-side `features` table:
```toml
[features]
personality = true
unified_exec = false
```
Only canonical feature keys are allowed in the requirements `features`
table; omitted keys remain unconstrained.
- Added a requirements-side pinned feature map to
`ConfigRequirementsToml`, threaded it through source-preserving
requirements merge and normalization in `codex-config`, and made the
TOML surface use `[features]` (while still accepting legacy
`[feature_requirements]` for compatibility).
- Exposed `featureRequirements` from `configRequirements/read`,
regenerated the JSON/TypeScript schema artifacts, and updated the
app-server README.
- Wrapped the effective feature set in `ManagedFeatures`, backed by
`ConstrainedWithSource<Features>`, and changed its API to mirror
`Constrained<T>`: `can_set(...)`, `set(...) -> ConstraintResult<()>`,
and result-returning `enable` / `disable` / `set_enabled` helpers.
- Removed the legacy-usage and bulk-map passthroughs from
`ManagedFeatures`; callers that need those behaviors now mutate a plain
`Features` value and reapply it through `set(...)`, so the constrained
wrapper remains the enforcement boundary.
- Removed the production loophole for constructing unconstrained
`ManagedFeatures`. Non-test code now creates it through the configured
feature-loading path, and `impl From<Features> for ManagedFeatures` is
restricted to `#[cfg(test)]`.
- Rejected legacy feature aliases in enterprise feature requirements,
and return a load error when a pinned combination cannot survive
dependency normalization.
- Validated config writes against enterprise feature requirements before
persisting changes, including explicit conflicting writes and
profile-specific feature states that normalize into invalid
combinations.
- Updated runtime and TUI feature-toggle paths to use the constrained
setter API and to persist or apply the effective post-constraint value
rather than the requested value.
- Updated the `core_test_support` Bazel target to include the bundled
core model-catalog fixtures in its runtime data, so helper code that
resolves `core/models.json` through runfiles works in remote Bazel test
environments.
- Renamed the core config test coverage to emphasize that effective
feature values are normalized at runtime, while conflicting persisted
config writes are rejected.
- Ran `compact_resume_after_second_compaction_preserves_history` inside
an explicit 8 MiB test thread and Tokio runtime worker stack, following
the existing larger-stack integration-test pattern, to keep the Windows
`compact_resume_fork` test slice from aborting while a parallel
investigation continues into whether some of the underlying async
futures should be boxed.
## Verification
- `cargo test -p codex-config`
- `cargo test -p codex-core feature_requirements_ -- --nocapture`
- `cargo test -p codex-core
load_requirements_toml_produces_expected_constraints -- --nocapture`
- `cargo test -p codex-core
compact_resume_after_second_compaction_preserves_history -- --nocapture`
- `cargo test -p codex-core compact_resume_fork -- --nocapture`
- Re-ran the built `codex-core` `tests/all` binary with
`RUST_MIN_STACK=262144` for
`compact_resume_after_second_compaction_preserves_history` to confirm
the explicit-stack harness fixes the deterministic low-stack repro.
- `cargo test -p codex-core`
- This still fails locally in unrelated integration areas that expect
the `codex` / `test_stdio_server` binaries or hit existing `search_tool`
wiremock mismatches.
## Docs
`developers.openai.com/codex` should document the requirements-side
`[features]` table for enterprise and MDM-managed configuration,
including that it only accepts canonical feature keys and that
conflicting config writes are rejected.
load plugin-apps from `.app.json`.
make apps runtime-mentionable iff `codex_apps` MCP actually exposes
tools for that `connector_id`.
if the app isn't available, it's filtered out of runtime connector set,
so no tools are added and no app-mentions resolve.
right now we don't have a clean cli-side error for an app not being
installed. can look at this after.
### Tests
Added tests, tested locally that using a plugin that bundles an app
picks up the app.
Update config.toml plugin entries to use
<plugin_name>@<marketplace_name> as the key.
Plugin now stays in
[plugins/cache/marketplace-name/plugin-name/$version/]
Clean up the plugin code structure.
Add plugin install functionality (not used yet).
Support loading plugins.
Plugins can now be enabled via [plugins.<name>] in config.toml. They are
loaded as first-class entities through PluginsManager, and their default
skills/ and .mcp.json contributions are integrated into the existing
skills and MCP flows.
sayan-oai ·
pakrym-oai ·
felixxia-oai ·
Dylan Hurd ·
Chris Dong ·
jif ·
Adam Perry @ OpenAI ·
Eric Traut ·
Matthew Zeng ·
starr-openai ·
Ahmed Ibrahim ·
neil-oai ·
Michael Bolin ·
rhan-oai ·
alexsong-oai ·
Anton Panasenko ·
xl-openai ·