## Summary
- surface shared rollout-budget exhaustion as
`CodexErr::RolloutBudgetExceeded` instead of a generic interrupted turn
- map it through the existing `CodexErrorInfo` and app-server v2
`codexErrorInfo` path
- keep local compaction from retrying after the shared rollout budget is
exhausted
This gives app-server clients a stable `rolloutBudgetExceeded` error
they can classify without guessing from `status="interrupted"`.
## Tests
- `just test -p codex-core rollout_budget`
## Summary
- add validated protocol-version, capability, and session identifier
types
- define explicit `ClientToHost` and `HostToClient` JSON envelopes for
connection negotiation and session open/close acknowledgements
- reject invalid states and unknown fields during decoding, with
explicit wire-format and round-trip coverage
## Why
This establishes the transport-neutral encoding shape needed to build
and test the new code-mode host incrementally. Cell, tool callback, and
failure-domain messages are intentionally deferred until their actors
and behavior tests establish the required semantics.
This is additive protocol scaffolding and does not change the current
production code-mode implementation.
## Validation
## Why
Selected capability roots belong to the executor filesystem, not the
app-server host. Converting their path strings into the host's native
`Path` breaks whenever the two machines use different path conventions,
such as a Windows executor behind a Unix app-server.
This PR establishes `PathUri` as the selected-plugin boundary so the
executor remains authoritative for its paths.
## What changed
- Require `selectedCapabilityRoots[].location.path` to be a canonical
`file:` URI and deserialize it directly as `PathUri`; native path
strings are rejected.
- Update the app-server schema, generated TypeScript, examples, and
request coverage for the URI contract.
- Keep selected roots, resolved plugin locations, manifest paths, and
manifest resources as `PathUri`.
- Inspect and read plugin roots and manifests only through the selected
environment's `ExecutorFileSystem`.
- Parse executor manifests with the shared URI-native parser from #29620
instead of projecting them onto the host filesystem.
- Enforce resource containment lexically and preserve the root URI's
POSIX or Windows path convention.
- Cover foreign Windows plugin roots and URI-native manifest resources.
```text
thread/start
selectedCapabilityRoots[].location.path = "file:///C:/plugins/demo"
| PathUri
v
ExecutorFileSystem
|
+--> plugin.json
+--> manifest resources
```
This PR stops at the shared selected-plugin representation. The next two
PRs remove the remaining host-path projections in the skill and MCP
consumers.
## Stack
1. #29614 — add lexical `PathUri` containment.
2. #29620 — share URI-native manifest path resolution.
3. **This PR** — keep selected plugin roots and resources URI-native.
4. #29626 — load executor skills without host path conversion.
5. #29628 — resolve executor MCP working directories without host path
conversion.
## Why
PR #29494 made context-window IDs visible to the model by wrapping the
token-budget window payload in `<context_window>`, but rollout JSONL
consumers still could not see the initial window identity by tailing the
session file. Compacted rollout items carry window IDs only after
compaction has happened, so a session with no compaction had no durable
JSONL record for window 0.
This change gives tailing consumers a stable initial-window record at
session creation time.
## What Changed
- Added `session_meta.context_window.window_id` for the initial
context-window identity.
- `CreateThreadParams` now requires `initial_window_id: String`, so
thread-store callers cannot accidentally create new threads without
window-0 metadata.
- Live thread creation derives the persisted initial window ID from the
same `AutoCompactWindowIds` used to initialize `SessionState`, keeping
runtime state and JSONL metadata aligned.
- Rollout reconstruction uses `session_meta.context_window.window_id` as
the initial-window fallback and derives `window_number = 0`,
`first_window_id = window_id`, and `previous_window_id = None`
internally.
- Fork reconstruction intentionally uses the same rollout reconstruction
path; consumers that need to distinguish copied initial-window metadata
can use the rollout `thread_id`.
- Legacy compactions without `window_number` still use compaction-count
fallback accounting instead of being reset to window 0 by the
initial-window fallback.
- Compacted rollout metadata still takes precedence once compaction
records exist, preserving the richer chain fields there.
## JSONL Shape
Real rollout JSONL is one object per line. This example is expanded for
readability, but shows the new initial `session_meta.context_window`
record followed by the existing compacted rollout item shape that also
carries window IDs:
```jsonl
{
"timestamp": "2026-06-22T12:00:00.000Z",
"type": "session_meta",
"payload": {
"session_id": "<THREAD_ID>",
"id": "<THREAD_ID>",
"timestamp": "2026-06-22T12:00:00.000Z",
"cwd": "/repo",
"originator": "codex",
"cli_version": "0.0.0",
"source": "cli",
"model_provider": "<MODEL_PROVIDER>",
"context_window": {
"window_id": "<INITIAL_WINDOW_ID>"
}
}
}
...
{
"timestamp": "2026-06-22T12:34:56.000Z",
"type": "compacted",
"payload": {
"message": "<COMPACTION_SUMMARY>",
"replacement_history": [
"..."
],
"window_number": 1,
"first_window_id": "<INITIAL_WINDOW_ID>",
"previous_window_id": "<INITIAL_WINDOW_ID>",
"window_id": "<NEXT_WINDOW_ID>"
}
}
```
The nested `context_window` object is intentional: it gives rollout
consumers a stable namespace for context-window metadata while only
writing the non-derivable initial `window_id`. For the initial window,
`window_number`, `first_window_id`, and `previous_window_id` are derived
internally instead of being written to the rollout.
## Verification
- `just test -p codex-protocol`
- `just test -p codex-rollout
recorder_materializes_on_flush_with_pending_items`
- `just test -p codex-core reconstruct_history`
- `just test -p codex-core
record_initial_history_reconstructs_forked_transcript`
- `just test -p codex-thread-store`
- `just test -p codex-state`
- `just test -p codex-app-server
thread_read_returns_summary_without_turns`
- `just test -p codex-rollout persistence_metrics`
## Why
I'd originally added `PathUri` legacy path deserialization thinking we'd
want it for having `PathUri` in public app-server APIs. Since then we've
added `LegacyAppPathString` to handle the messy conversions that we need
for backcompat. It's confusing for `PathUri` to support deserializing
legacy paths when we don't yet want to actually expose app-server
callers or rollout storage to the new URI format.
Stacked on top of #29472 to avoid breaking compatibility in case those
types ended up stored somewhere for someone.
## What changed
- Parse deserialized `PathUri` values exclusively as valid `file:` URIs.
- Replace legacy acceptance coverage with rejection coverage for
top-level filesystem paths and sandbox working directories.
- Serialize CWDs in hand-built exec-server process requests as `PathUri`
values.
## Why
Use a clearer name for what happens when this helper sets up a test
environment.
## What
- Rename the builder and its harness wrapper to use `auto_env` instead
of `remote_env` because the helper will set up a local environment if
configured by the build system.
## Why
Core tests should branch on the executor's operating system, not on
runner details such as Docker or Wine. This keeps platform behavior
stable as new test backends are added and reserves Wine-specific skips
for actual runner debt.
## What
- Add `TestTargetOs` and target/host-aware skip helpers while keeping
`TestEnvironment` internal.
- Replace topology enum access with remote predicates and a narrow
Docker accessor.
- Migrate OS-semantic Wine skips, preserve runner-specific gaps, and
document the skip taxonomy.
## Validation
- `just test -p core_test_support`
- `just test -p codex-core
remote_test_env_can_connect_and_use_filesystem`
- `bazel test //codex-rs/core:core-all-wine-exec-test
--test_output=errors` reached test execution; unrelated existing
view-image, path, and timing failures remain.
- `just test -p codex-core` and `just test` reached broad test
execution; this checkout has unrelated helper, sandbox, and timing
failures.
Mechanical rename of CodeModeService => InProcessCodeModeSession
This already implements a CodeModeSession as its prime interface to
Core. The name was vestigial _and_ confusing af when embedded inside
core::tools::code_mode::CodeModeService
## Description
Rename the experimental app-server item pagination API from
`thread/turns/items/list` to `thread/items/list` and make `turnId`
optional. Clients can now page persisted items across a thread, or still
filter to one turn when needed.
## What changed
- Rename the request/response protocol types and JSON-RPC method to
`ThreadItemsList*` / `thread/items/list`.
- Pass optional `turnId` through to `ThreadStore::list_items`.
- Update app-server docs and focused protocol/app-server tests.
## Validation
- `just test -p codex-app-server-protocol thread_items_list_round_trips`
- `just test -p codex-app-server thread_items_list_returns_unsupported`
## Summary
- add the exec-server working directory to `environment/info` as an
optional `PathUri`
- populate it from the executor process's current directory
- preserve compatibility with older responses that omit `cwd`
## Why
Remote clients currently have no executor-native default working
directory. This forces callers such as app-server-backend to assume
`/workspace`, which fails for laptop environments. Reporting the cwd
alongside the detected shell lets clients use the path convention and
location of the actual executor.
## Impact
This is backward-compatible: the new response field is optional, and
clients can continue handling responses from older exec servers. A
follow-up app-server-backend change will consume the value for cwd-less
`command/exec` requests.
## Validation
- `just test -p codex-exec-server` (275 passed, 2 skipped)
## Why
Plugin manifests use the same schema whether the package lives on the
host or in an executor. Only the path representation differs: host
callers need native `Path` inputs and `AbsolutePathBuf` outputs, while
executor callers need `PathUri` throughout.
Maintaining separate parsing or resolver implementations would duplicate
the manifest rules and allow them to drift. This PR instead makes
URI-native resolution the single parsing path and keeps host conversion
at the boundary.
## What changed
- Make `parse_plugin_manifest_uri` the shared manifest parser and
resolve every path-bearing field as `PathUri`.
- Keep the existing host entrypoint as a thin adapter: convert its
native root and manifest path to `PathUri`, run the shared parser, then
map resources back to `AbsolutePathBuf`.
- Expose `PluginManifest::try_map_resources` so callers can convert the
generic resource type without duplicating manifest construction.
- Resolve relative manifest paths using the root URI's convention:
backslashes are separators for Windows roots and ordinary filename
characters for POSIX roots.
- Apply lexical containment after URI resolution, rejecting absolute
paths and parent traversal outside the plugin root.
- Make encoded backslashes fail containment only for Windows URIs;
encoded `/` remains unsafe for every convention.
- Use a host-native synthetic root for marketplace fallback manifests so
the host adapter also works on Windows.
```text
host Path --------> PathUri --\
+--> one manifest parser --> PluginManifest<PathUri>
executor PathUri -------------/
host result: PluginManifest<PathUri> --> PluginManifest<AbsolutePathBuf>
```
Existing host manifest behavior is preserved; #28918 is the first
executor consumer.
## Verification
- `just test -p codex-utils-path-uri`
- `just test -p codex-plugin`
- `just test -p codex-core-plugins`
## Stack
1. #29614 — add lexical `PathUri` containment.
2. **This PR** — share URI-native manifest path resolution.
3. #28918 — keep selected plugin roots and resources URI-native.
4. #29626 — load executor skills without host path conversion.
5. #29628 — resolve executor MCP working directories without host path
conversion.
## Why
auto review reviews Codex App tool calls using connector metadata such
as the app ID, name, and description. That metadata does not identify
the account behind the OAuth connection.
For Google Drive, this means auto review cannot distinguish a Drive
connection authenticated as `user@email.com` from a personal Drive
account. Uploading work data can therefore look like a transfer to a
personal destination even though the connector service already knows the
authenticated account email.
## What changed
- Read `_meta._codex_apps.connected_account_email` while resolving
approval metadata for built-in Codex App tools.
- Include the connected account email in the structured MCP tool action
sent to auto review.
- Trim empty values and omit the field when the connector link has no
account email.
- Update existing auto review request constructors and add coverage for
request construction and JSON serialization.
## Security
Only metadata from the trusted built-in `codex_apps` MCP server is
accepted. Custom MCP servers cannot inject a connected account email
into auto review reviews; the new regression test verifies that spoofed
metadata is ignored.
The email is used only in auto review's private review request. This
change does not add it to model-visible tool descriptions, app-server
approval events, or auto review assessment/review analytics.
[Codex Thread
019edc37-5345-7272-92c9-bf5494cf3819](https://codex-thread-link.openai.chatgpt-team.site/thread/019edc37-5345-7272-92c9-bf5494cf3819)
## Summary
- count MCP `CallToolResult.isError` responses as failed calls instead
of successful transport-level calls
- add `codex.mcp.call.error` with bounded `error_type` and trusted
plugin-service `error_code` dimensions
- record the same error classification on MCP tool-call spans while
keeping untrusted server error text out of metric labels
## Scope
- no changes to MCP routing, retries, tool behavior, configuration, or
public APIs
- request failures remain grouped as `mcp_request`; separating
connection, timeout, protocol, and JSON-RPC failures requires preserving
typed errors through the existing flattened error boundary
## Testing
- `just test -p codex-core 'mcp_tool_call::tests::'` (75 passed)
- `just fix -p codex-core`
- `just fmt`
- `just test -p codex-core` (2,676 passed; 80 unrelated environment
failures from missing test binaries, sandbox signals, and read-only
paths)
## Why
With deferred executors, an environment can become ready between two
sampling requests in the same turn. The model-visible environment
update, advertised tools, and eventual tool execution must all describe
the same request-time view.
Otherwise, a request built while only environment B is ready can
advertise a tool without an `environment_id`; if higher-priority
environment A becomes ready before execution, that call could silently
run in A instead.
This PR is stacked on #29527.
## Design
`run_turn` captures one `Arc<StepContext>` at each sampling-request
boundary. That step owns the request's `TurnContext` and environment
snapshot.
- World-state environment updates and tool planning borrow that same
step.
- `ToolCallRuntime` retains the `Arc` while asynchronous tool calls
execute.
- `ToolInvocation` carries the step to handlers; its temporary `turn`
compatibility field is derived from the same object.
- `ToolRouter` does not retain `StepContext`; it only uses it while
constructing the request's tool set.
- With `DeferredExecutor` disabled, step capture keeps using the
environments frozen at turn start.
Simply: every sampling request gets one consistent picture of its
environments, from what the model sees through where its tool calls run.
## What changed
- Build environment-dependent tool specs from the current request's
`StepContext`.
- Use that same step for unified exec, legacy shell, `apply_patch`,
`view_image`, and `request_permissions` execution.
- Hide environment-backed tools, including `request_permissions`, while
no environment is attached.
- Resolve legacy shell paths and metadata from the selected step
environment instead of the stale turn-start environment.
- Capture explicit steps at non-turn-loop boundaries such as compaction,
prompt debug, and startup prewarm.
- Reconcile prompt-debug history from the same step used to build its
tools.
## Follow-up
- Bind yielded code-mode cells to the tool runtime that created them, so
nested calls made after yielding continue to use the originating
request's `StepContext`.
## Test plan
- `just test -p codex-core
deferred_executor_updates_context_and_tools_after_startup`
- `just test -p codex-core
environment_count_controls_environment_backed_tools`
- `just test -p codex-core
build_prompt_input_includes_context_and_user_message`
## Summary
- replace the stale `AskForApproval::OnFailure` reference in the MCP
connection manager test with `AskForApproval::OnRequest`
- restore `codex-mcp` test compilation after `OnFailure` was removed in
#28418
## Root cause
The test was added on main after the approval-policy removal branch had
already updated the other references, so the newly added call site was
missed when #28418 merged.
## Validation
- `just test -p codex-mcp` (90 passed)
- `just fmt`
## Why
view_image needs to support foreign OS remote executors.
## What
- resolve image paths against the selected environment as `PathUri` and
read them through that environment's filesystem
- keep app-server's public path field wire-compatible as
`LegacyAppPathString`, with purpose-specific UI rendering
- cover relative and absolute target-native paths in the core
integration test and run the full `view_image` suite under wine-exec
without skips
## Summary
- allow the native Responses API `image_generation` tool when the active
provider carries CCA's non-empty `x-openai-actor-authorization` header
- preserve the Codex-managed ChatGPT auth path, scoped to providers that
actually require OpenAI auth
- keep generic custom providers excluded, including when unrelated
ChatGPT credentials are cached
- retain the existing feature, provider-capability, and
image-input-modality gates
## Why
CCA authenticates its inference requests through the active provider's
`x-openai-actor-authorization` and `ChatGPT-Account-ID` headers, so it
does not have a Codex-managed login session. The previous gate therefore
hid the native hosted image-generation tool despite an authenticated
codex-backend path.
This change is intentionally limited to the native hosted tool. It adds
no extension, MCP, plugin-service, session-source, token plumbing, or
new provider configuration surface.
## Tests
- `cargo test -p codex-core
hosted_tools_follow_provider_auth_model_and_config_gates`
- `cargo fmt --all -- --check`
- `git diff --check origin/main`
## Why
Filesystem helpers intentionally run with a minimal environment that
excludes proxy variables. After filesystem operations started using the
Windows sandbox wrapper, the wrapper derived an empty proxy
configuration from that helper environment and compared it with the
persistent sandbox setup marker. When the marker contained proxy ports,
every filesystem operation appeared to require a firewall update, which
could launch elevated setup, show a UAC or loader dialog, and fail
operations such as `apply_patch` with error 1223.
Filesystem helpers do not use network access, so they should preserve
the proxy/firewall state established by normal sandboxed process
launches.
## What changed
- Add an explicit Windows sandbox proxy-settings mode for reconciling or
preserving persistent proxy state.
- Use preserve mode for filesystem helpers while normal process launches
continue to reconcile proxy settings from their environment.
- Carry the selected proxy state consistently through setup validation,
elevated setup, and non-elevated ACL refreshes.
- Cover wrapper argument propagation and marker-derived proxy
preservation.
## Validation
- `cargo build -p codex-cli --bin codex`
- `just test -p codex-windows-sandbox
preserving_proxy_settings_uses_the_existing_marker`
- `just test -p codex-windows-sandbox windows_wrapper_args_round_trip`
- `just test -p codex-windows-sandbox
setup_request_prefers_explicit_proxy_settings`
- `just test -p codex-sandboxing transform_for_direct_spawn_windows`
- `just test -p codex-exec-server fs_sandbox::tests`
- Ran the same sandboxed `fs/writeFile` reproduction against published
`0.142.0-alpha.6` and the new CLI. The published CLI launched elevated
setup and failed with `ShellExecuteExW ... 1223`; the new CLI completed
without elevation.
Related to #28359.
## Why
Plugin analytics overloaded `plugin_id`: most events used the Codex
`<plugin>@<marketplace>` identity, while remote install events used the
backend plugin ID. That makes the same field change meaning across event
types and complicates downstream identity resolution.
This change makes the contract unambiguous:
- `plugin_id`: the local Codex `<plugin>@<marketplace>` identity, when
resolved
- `remote_plugin_id`: the backend plugin identity, when available
For a remote install failure that happens before plugin details resolve,
`plugin_id` is `null` and `remote_plugin_id` remains populated.
## What changed
All six plugin analytics events use the same identity contract:
- `codex_plugin_installed`
- `codex_plugin_install_failed`
- `codex_plugin_uninstalled`
- `codex_plugin_enabled`
- `codex_plugin_disabled`
- `codex_plugin_used`
Remote identity is resolved from the current installed-plugin snapshot
first, with persisted install metadata as fallback. The telemetry
metadata type keeps local identity optional for failures that occur
before remote details are available.
The app-server test client's manual analytics smokes now find remote
mutation events through `remote_plugin_id` and validate that `plugin_id`
remains local.
## Remote uninstall
Resolve and capture telemetry metadata before removing the local plugin
cache, then emit `codex_plugin_uninstalled` after the backend confirms
success. The event is also emitted when backend uninstall succeeds but
local cache cleanup reports `CacheRemove`.
If a concurrent remote-cache refresh removes the local bundle before
telemetry capture, the already-fetched remote plugin detail supplies
fallback capability metadata.
## Validation
- `just test -p codex-analytics` — 82 passed
- `just test -p codex-core-plugins` — 271 passed
- `just test -p codex-app-server-test-client` — 5 passed
- `just test -p codex-plugin` — 3 passed
- `just test -p codex-app-server plugin_install` — 37 passed
- `just test -p codex-app-server plugin_uninstall` — 10 passed
The production app-server install/uninstall flow was also exercised
against `plugins~Plugin_f1b845ac33888191ac156169c58733c2`
(`build-ios-apps@openai-curated-remote`), and the plugin's original
uninstalled state was restored.
## Why
The managed MITM trust bundle must be readable by sandboxed commands.
Persisting its sibling CA private key under `$CODEX_HOME/proxy`
therefore requires a deny-read sandbox rule, but the Windows unelevated
backend rejects deny-read paths and WSL1's legacy Landlock path cannot
enforce that rule.
A persistent OS credential store also does not provide the same
cross-platform boundary from other processes running as the same user.
Keeping the signer inside the network proxy process avoids both
problems: ordinary sandbox setup stays independent of CA-key state, and
no private signing key is exposed through the filesystem or a persistent
credential record.
## What
- generate one managed CA per proxy process and retain its private
signer only in proxy memory
- emit only content-addressed public CA certificates and trust bundles
under `$CODEX_HOME/proxy`
- hold a cross-process lease for each active public certificate and
prune artifacts from inactive proxy processes
- keep all CA ownership in `codex-network-proxy`; no `codex-core` or
sandbox-policy changes
- validate generated trust-bundle paths by their content hash
- keep the public bundle readable by sandboxed commands on Windows,
WSL1, macOS, and Linux
The independent startup custom-CA follow-up is #29014.
## Validation
- `CODEX_HOME=/private/tmp/codex-test-home-network-proxy just test -p
codex-network-proxy` (179 tests)
- `just bazel-lock-check`
- `just fix -p codex-network-proxy`
- `just fmt`
---------
Co-authored-by: viyatb-oai <viyatb@openai.com>
## Why
Managed network configures commands to use local HTTP and SOCKS proxies.
For commands delegated to the exec server, the proxy environment and the
sandbox policy were prepared separately. On macOS, that meant a command
could receive `HTTPS_PROXY=http://127.0.0.1:43123` while Seatbelt still
denied access to port `43123`.
## What changed
`NetworkProxy` now prepares the command environment and sandbox context
together from the same runtime snapshot:
```text
Prepared managed network
├── command environment: HTTPS_PROXY=http://127.0.0.1:43123
└── sandbox context: allow outbound to 127.0.0.1:43123
```
That context travels with remote exec requests. The exec server
preserves the managed proxy and CA environment, and macOS Seatbelt
allows only the prepared loopback proxy ports without enabling broad
network access or local binding.
The protocol field is optional and the existing enforcement flag remains
in place, preserving compatibility with callers that do not send the new
context.
## Why
The rollout persistence metrics added on current `main` exhaustively
match `ResponseItem`, but omit `ResponseItem::AdditionalTools`. That
prevents `codex-rollout` and downstream targets from compiling across
Cargo and Bazel builds.
## What
Map `ResponseItem::AdditionalTools` to the `response.additional_tools`
metric label, consistent with the existing exact-variant labels.
## Validation
- `just test -p codex-rollout` (76 passed)
- `just fix -p codex-rollout`
## Summary
Handle `ResponseItem::AdditionalTools` in rollout persistence metrics.
The persistence metrics match was added after the `AdditionalTools`
variant and omitted it, causing release builds to fail with a
non-exhaustive pattern error. This assigns the item the
`response.additional_tools` metrics label.
Release failure:
https://github.com/openai/codex/actions/runs/28043786727/job/83016608475
## Validation
- `just fmt`
- `just test -p codex-rollout` (76 passed)
## Why
Follow-up to #29249 and its [compaction review
thread](https://github.com/openai/codex/pull/29249#discussion_r3455055101).
During a turn, environment readiness can change between sampling
requests. Inline compaction must render the same model-visible
`WorldState` used by the request it follows. Rebuilding that state
during compaction can observe a newer environment, make replacement
history disagree with what the model saw, and suppress the next
environment update.
## What changed
- Make `run_turn` own the current `Arc<WorldState>` and replace it only
between sampling requests.
- Build each state from an explicitly chosen environment snapshot, diff
deferred-executor steps against the turn-owned state, and retain the
latest state in `ContextManager` only for cross-turn and resume
tracking.
- Pass the exact turn-owned state into inline compaction and explicit
new-context-window replacement.
- Carry that state with
`InitialContextInjection::BeforeLastUserMessage`, so replacement context
and its stored baseline cannot come from different snapshots.
- Remove obsolete state-recapture helpers and ambiguous TurnContext-only
WorldState builders.
- Add an integration test that moves an environment from starting to
ready during a paused turn, triggers compaction, and verifies the next
request receives the readiness update exactly once.
## Test plan
- `just test -p codex-core
deferred_executor_compaction_preserves_then_updates_environment_once`
- `just test -p codex-core process_compacted_history`
- `just test -p codex-core mid_turn_continuation_compaction`
- `just test -p codex-core build_initial_context`
- `just test -p codex-core
ignores_session_prefix_messages_when_truncating`
## Summary
MCP refresh replaced the published connection manager without shutting
down the manager it superseded. If another task retained that old
manager, its stdio MCP processes stayed alive and accumulated across
refreshes.
Atomically swap in the refreshed manager, then explicitly shut down the
exact manager returned by the swap. Add a process-level regression test
that retains the old manager during refresh and verifies its stdio
process exits while the replacement remains available.
## Context
Explicit cleanup was lost when manager publication moved to `ArcSwap`.
Dropping the old manager is not a reliable shutdown boundary because
active callers can retain its `Arc` and underlying client process
handles.
## Summary
- rename `reminder_interval_model_requests` to
`reminder_interval_seconds`
- read the configured time provider before every model request and
inject a reminder only after the configured number of seconds has
elapsed
- preserve immediate first delivery and forced delivery after compaction
changes the context window
## Tests
- `just test -p codex-core current_time_reminder`
- Add 1%-sampled rollout persistence metrics that report per-item and
per-thread JSON byte totals before and after filtering when metrics
export is enabled.
- Tag each item with its exact response or event variant, including
nested turn-item kinds for conditionally persisted completion events, so
aggregate cloud-storage impact can be estimated by policy choice.
## Summary
- Update `rmcp` and `rmcp-macros` from 1.7.0 to 1.8.0.
- Adapt to the new shared `peer_info` return type.
- Box OAuth status discovery at the MCP boundary to keep the expanded
future type from overflowing Rust's trait recursion limit.
This brings in custom OAuth HTTP client support from
[modelcontextprotocol/rust-sdk#908](https://github.com/modelcontextprotocol/rust-sdk/pull/908).
## Summary
Resuming a persisted thread currently deep-clones its complete rollout
history several times. `InitialHistory` is retained for the app-server
response, copied into thread persistence, and copied again by read-only
accessors. These copies scale with the complete rollout rather than the
bounded model context and add measurable latency for large sessions.
This change stores resumed rollout history in `Arc<Vec<RolloutItem>>`.
Rollout loading wraps the parsed vector once, while app-server response
construction, session initialization, and thread persistence share it
through inexpensive `Arc` clones. Read-only history access now returns a
borrowed slice, and fork paths use `Arc::unwrap_or_clone` where they
genuinely need mutable ownership. Rollout reconstruction also consumes
its temporary context instead of cloning the reconstructed model
history.
The serialized representation remains unchanged. In an artificial 123 MB
rollout benchmark, sharing resumed history reduced cold resume latency
by roughly 9–10%. The affected crates compile with their test targets,
all 80 thread-store tests pass, and the Bazel dependency lock remains
valid.
## Why
Executor-owned paths must stay portable while the orchestrator reasons
about them. Converting a Windows or remote path to the orchestrator
host's native path just to check containment breaks that boundary.
## What changed
- Add lexical containment to `PathUri`.
- Compare URI authorities and complete path segments, so `plugin-other`
is not treated as a child of `plugin`.
- Fail closed for encoded path separators and opaque fallback URIs.
For example:
```text
file:///C:/plugins/foo/assets/icon.svg
is below file:///C:/plugins/foo
file:///C:/plugins/foo2/icon.svg
is not below file:///C:/plugins/foo
```
This is the shared foundation for keeping executor-owned plugin
resources URI-native without consulting the orchestrator filesystem.
## Summary
Multi-agent v2 tools now use the fixed `collaboration` namespace when
namespace tools are available. This keeps the model-visible hint and the
actual tool surface aligned around `functions.collaboration.*`, without
exposing an unshipped namespace knob to users.
The PR also removes the old `features.multi_agent_v2.tool_namespace`
config/schema surface, updates the MAv2 test fixtures for namespaced
calls, and fixes stale `TurnContext.features` references that were
breaking `codex-core` builds.
## Changes
- Expose MAv2 tools under `collaboration` instead of relying on a
configurable namespace.
- Remove `tool_namespace` from MAv2 TOML config, resolved config,
validation, schema, and tests.
- Update tool-planning and integration fixtures to assert or emit
namespaced MAv2 tool calls.
- Read feature state through `TurnContext.config.features` in the
multi-agent mode context paths.
## Testing
- `just write-config-schema`
- `just test -p codex-features`
## Summary
- Require the reserved Codex Apps MCP server name to be present in the
connection manager before treating it as host-owned.
- Update auth elicitation tests to model an installed host-owned Codex
Apps server without sending startup events to the test session.
## Why
PR #29518 replaced the old host-owned flag with a name-only check. That
made non-host-owned tests with the reserved codex_apps name enter auth
elicitation and wait forever for a response.
## Why
The 0.142.0 persistent-log filter disables target=log, but bridged log
records are filtered using their original dependency target before
tracing-log emits them as target=log. This allowed high-volume
dependency TRACE events to keep reaching SQLite.
This is a follow-up to #28224.
## What changed
- Reject bridged target=log events inside the SQLite sink before
formatting or queueing them.
## Summary
- let `codex sandbox` accept the JSON value from
`codex/sandbox-state-meta`
- require the payload `permissionProfile` instead of falling back to
ambient permissions
- reuse the existing macOS, Linux, and Windows launch paths, treating
external sandbox state conservatively as read-only
- let opaque forwarders add runtime read roots and disable direct
network access without decoding the payload
Builds on #29113, which is now on `main`.
## Tests
- `just test -p codex-cli debug_sandbox::tests`
- `cargo build -p codex-rmcp-client --bin test_stdio_server`
- `just test -p codex-core
stdio_mcp_tool_call_includes_sandbox_state_meta`
- `just test -p codex-mcp`
- `just fmt`
## Why
`McpConnectionManager::new` classified the Codex Apps server twice: once
to create its tools cache context and again to select its runtime
authentication provider. Keeping those decisions separate makes it
harder to see that they belong to the same server-specific setup path.
## What changed
- Group Codex Apps cache and authentication setup under one explicit
branch.
- Keep regular MCP server setup in the corresponding `else` branch.
- Limit environment bearer-token inspection to the Codex Apps path where
it affects runtime authentication.
## Why
Codex Apps cache writes are already restricted to Codex Apps call paths:
startup invokes the helper only from the Codex Apps branch, and hard
refresh operates on the reserved Codex Apps server directly. Rechecking
the server name inside the cache helper duplicates that classification
and leaves the helper with an argument that cannot change valid
behavior.
## What changed
- Remove the redundant server-name check and parameter from the cache
writer.
- Rename the helper to `write_codex_apps_tools_cache` to reflect its
narrower contract.
- Update production and test callsites to use the simplified API.
## Why
`ResponseItem::AdditionalTools` was added without updating app-server
image URL validation. The exhaustive match therefore prevents app-server
and downstream targets from compiling on `main`.
## What changed
Treat `AdditionalTools` like the other response items that cannot
contain input-image URLs.
## Why
Codex Apps-specific behavior is currently distributed across cache
helpers, startup, tool conversion, and model-visible annotation. Each
layer independently checks the reserved server name, which obscures the
boundary between trusted host-owned connector metadata and regular MCP
server data.
Classifying the server once when `AsyncManagedClient` is created gives
the client a single source of truth and makes the two processing paths
explicit.
## What changed
- Record whether an `AsyncManagedClient` represents the Codex Apps
server at construction time.
- Route startup cache loading, cache persistence, and cache telemetry
through the Codex Apps branch.
- Split uncached tool conversion between Codex Apps normalization and
regular MCP metadata sanitization.
- Split model-visible schema and plugin provenance handling along the
same boundary.
- Remove redundant server-name guards from helpers that are now called
only from the Codex Apps branch.
## Verification
- Preserve behavioral coverage that verifies Codex Apps connector
metadata and the complete converted `ToolInfo` shape.
## Stack
Depends on #29518.
When using Responses Lite, we should all use `additional_tools` and a
developer item instead of the top level tools array & instructions
field. This keeps things 1-to-1.
Forced namespacing for _all_ tools will land in a following PR after
some coordination & fixes in Responses API (around collisions & return
items).
The goal is to eventually expand the scope of this to _all_ requests
from codex, but that will require larger coordination across providers &
slower rollout.
## Why
Codex Apps server admission is already decided before
`McpConnectionManager` is constructed. `effective_mcp_servers` and
`effective_mcp_servers_from_configured` remove the server when the apps
feature or required authentication is unavailable, so storing the same
decision on the manager duplicates state that can drift from the
effective server map.
## What changed
- Remove `host_owned_codex_apps_enabled` from `McpConnectionManager` and
its constructor.
- Identify the host-owned Codex Apps server by its reserved server name
once it is present in the effective server map.
- Remove the now-unused flag calculations and constructor arguments from
production and test callsites.
## Summary
- express remote compaction result handling as an exhaustive match
- preserve the special `TurnAborted` path without emitting a generic
compaction error
- rely on the standard `test_codex` provider setup in the compaction
budget test
Follow-up to review feedback on #28707.
## Testing
- `just test -p codex-core
compaction_budget_exhaustion_aborts_without_error_or_retry`
- `just fmt`
## Summary
NVIDIA asked to measure Fast mode usage and reasoning effort from Codex
CLI OTEL logs. Add the finalized `service_tier` and
`model_reasoning_effort` to the existing `codex.sse_event`
`response.completed` record.
This intentionally reuses the existing completion event and leaves
transport APIs and shared telemetry plumbing unchanged.
## Testing
- `cargo build -p codex-cli --bin codex`
- `just test -p codex-core responses_api_emits_api_request_event`
- End-to-end with the built CLI and a local OTLP/HTTP collector:
- Fast/high emitted `service_tier=priority` and
`model_reasoning_effort=high` with token usage.
- Standard/low omitted `service_tier` and emitted
`model_reasoning_effort=low` with token usage.
## Why
On Windows, Codex uses a PowerShell safe-command classifier to decide
whether a command is read-only enough to run without additional
approval. The classifier lowers `EndBlock.Statements` into argv-like
command words and checks those words against a safelist.
PowerShell can execute code stored elsewhere in the AST. Parameter
defaults, named blocks, `using` preambles, and top-level `trap` handlers
are not represented in the lowered statement list. Ignoring those
regions can make a side-effecting script look like a read-only command.
## What
Fail closed whenever a PowerShell script contains executable AST content
that the current lowering does not represent.
## How
- Return `unsupported` for parameter, dynamic-parameter, begin, process,
and clean blocks.
- Return `unsupported` for `using module` and `using assembly`
preambles.
- Return `unsupported` for non-empty `EndBlock.Traps` collections.
- Preserve compatibility with Windows PowerShell 5.1 by looking up
`CleanBlock` dynamically.
- Treat `unsupported` as a failure to prove that the command is safe,
routing it through the normal approval path.
- Add parser-level and end-to-end regressions for parameter blocks,
named blocks, using statements, and trap handlers.
This does not make these PowerShell forms invalid or prevent them from
running. It prevents automatic safe-command approval when the classifier
cannot account for all executable behavior.
## Testing
- `just test -p codex-shell-command`
- Windows CI exercises the parser and end-to-end safe-command
regressions against a real PowerShell installation.
---------
Co-authored-by: viyatb-oai <viyatb@openai.com>
## Summary
- read the request-scoped safety-buffering treatment from HTTP response
headers and per-turn WebSocket metadata through one shared header parser
- combine that treatment with Responses API safety-buffering signals
- propagate `showBufferingUi` and nullable `fasterModel` through the
existing `model/safetyBuffering/updated` app-server notification
- update the app-server documentation and generated JSON and TypeScript
schemas
The public implementation contains no model mapping or real model
identifier. Tests and protocol examples use generic `current-model` and
`faster-model` placeholders only.
## Dependencies
- server-side treatment evaluation:
https://github.com/openai/openai/pull/1060247
- initial Responses API safety-buffering propagation:
https://github.com/openai/codex/pull/29371
- Codex App UI: https://github.com/openai/openai/pull/1057789
## Validation
- Codex API tests: 129 passed
- focused Codex core safety-buffering integration test passed
- app-server protocol tests passed after regenerating schema fixtures
- Clippy fix and repository formatting completed successfully
The broader app-server run compiled all changed crates and completed
with 1,269 passing tests. Its remaining failures were unrelated
environment limitations: macOS sandbox application was denied, one
expected test binary was unavailable, and several existing subprocess
tests timed out as a result.
## Why
Remote stdio MCP servers can run in an environment whose path convention
differs from the Codex host. A Windows cwd such as
`C:\Users\openai\share` is absolute for the executor but was rejected by
a POSIX orchestrator.
Built on #29501, now merged, which only clarifies the host-native
`PathUri` constructor name.
## What changed
- Deserialize MCP cwd values as `LegacyAppPathString` so config does not
apply host path rules.
- Interpret that spelling as host-native for local launches and convert
it to `PathUri` at executor launch.
- Skip host filesystem and command resolution checks for remote stdio in
`codex doctor`.
- Add host-independent config and executor-boundary coverage using the
foreign path convention for each test platform.
## Validation
- `just test -p codex-utils-path-uri -p codex-config -p codex-mcp -p
codex-rmcp-client` (408 passed)
- `just test -p codex-cli -p codex-rmcp-client` (372 passed)
- `cargo check --workspace --tests`
- `just test` (11,311 passed; 43 unrelated environment/timing failures)
- `just fix -p codex-cli -p codex-config -p codex-core -p codex-mcp -p
codex-mcp-extension -p codex-rmcp-client -p codex-tui`
rka-oaiandGitHub
Channing CongerandGitHub
jifandGitHub
Michael BolinandGitHub
Adam Perry @ OpenAIandGitHub
Owen LinandGitHub
Rasmus RygaardandGitHub
viyatb-oaiandGitHub
stevenlee-oaiandGitHub
sayan-oaiandGitHub
richardopenaiandGitHub
iceweasel-oaiandGitHub
jameswt-oaiandGitHub
Boyang NiuandGitHub
Dylan HurdandGitHub
TomandGitHub
Charlie MarshandGitHub
Ahmed IbrahimandGitHub
daniel-oaiandGitHub
Francis ChalisseryandGitHub
Celia ChenandGitHub