## Why
An empty tool map does not distinguish a successfully returned empty catalog from a server whose startup or tool discovery failed.
## What changed
- Add nullable `toolsError` to each `mcpServerStatus/list` entry when no catalog is available because startup or tool discovery failed.
- Keep `toolsError` null for returned catalogs, including cached and empty catalogs, and continue returning healthy servers when another server fails.
- Update the protocol schemas and generated TypeScript type.
## Testing
- Extend the app-server status integration test with a broken MCP server and verify its error without losing the healthy server's inventory.
GitOrigin-RevId: 30b4ca4d9a45cec6998f5f626d3fddbee90a525d
## What changed
- Add the under-development `mcp_oauth_refresh_coordination` feature and pin its selected refresh mode to each MCP connection.
- Treat the refresh mode as part of OAuth connection identity so runtime and MCP configuration refreshes reconnect when the mode changes.
- Keep legacy refresh behavior in builds where coordination is unavailable and emit a warning when coordinated mode is selected.
## Testing
- Cover connection replacement when the feature is toggled through both runtime and MCP configuration refresh paths.
GitOrigin-RevId: 89596150f0338e6ede0e712badecea4450831766
## Why
Event subscriptions may need to outlive the MCP runtime associated with a task.
## What changed
- Add `McpEventStreamOpener`, which retains the hosted Apps MCP connection settings and opens each event stream with an independent client.
- Keep subscription cancellation tied to account changes and removal of the task's event server, including when cancellation ownership is forwarded across runtime replacement.
- Expose the opener and cancellation forwarding through `McpResourceClient`.
## Testing
Add an HTTP transport test that shuts down the runtime, refreshes credentials for the same account, and verifies that a new event stream connects with the current authorization and receives notifications.
GitOrigin-RevId: e102aa382e63121661ff0567f7523515e459922d
## Why
Plugin MCP trust decisions must describe the exact host-discovered server
registration. Looking up a plugin again by ID can resolve a different root.
## What changed
- Retain the host plugin root in MCP catalog attribution and connection identity.
- Reconnect an MCP server when its attributed plugin root changes.
- Pass the retained root through tool lifecycle attribution and use it directly
when building trusted tool context.
## Testing
Add coverage for catalog identity, connection reconciliation, lifecycle
attribution, and rejecting a different root associated with the same plugin ID.
GitOrigin-RevId: bc92743dd40ab44065a07c3ea972b4956ef8351f
## What changed
- Fetch ChatGPT account access and attach `cyber_trusted_access` under
`openai/entitlementContext` when an installed or selected plugin explicitly
requests it from a local, read-only stdio tool with no arguments.
- Reject caller-supplied entitlement context on direct MCP calls, and fall back
to an `unknown` status when access cannot be verified or the account changes
during the lookup.
- Count entitlement lookup time against the MCP tool-call timeout.
## Testing
- Cover eligible calls plus unowned, remote, HTTP, mutable, argument-bearing,
API-key, undeclared, direct, and account-switch cases in core and app-server
integration tests.
GitOrigin-RevId: 303b61585ef4530276c373d914a2d0ccef1f4ac2
## Why
MCP servers attached to executor environments must retain their owner's
permission profile instead of inheriting the thread-wide sandbox authority.
## What changed
- Resolve and capture a permission profile for each enabled MCP server when
publishing the runtime, and reject calls or elicitations when that authority
is unavailable.
- Use the captured server profile for tool approval decisions, elicitation, and
sandbox metadata, including after runtime refreshes.
- Materialize `:workspace_roots` from `PathUri` values so permissions preserve
the path convention of remote executor environments.
- Give threadless app discovery and resource reads an explicit default
permission profile.
## Testing
Added coverage for per-server elicitation authority, unresolved attachments,
runtime refreshes, restricted tool calls, and foreign-platform workspace roots.
GitOrigin-RevId: 6b188d4b08b29e9971ff7aa68a7785a0cdf9e394
## Why
Reusing an MCP connection that is still starting after its
`startup_timeout_sec` changes leaves the old startup budget in effect.
## What changed
- Track the effective startup timeout on each MCP connection.
- Reuse a pending startup only when its timeout still matches the current
configuration.
- Continue reusing ready connections when only the startup timeout changes.
## Testing
Added integration coverage that refreshes the timeout while one server is
still initializing and another is ready, verifying that only the pending
startup is replaced.
GitOrigin-RevId: 5ed71a57d3bf0dd6e227aacd16acd3a35bc37ecf
## Why
MCP inventory can be cached or collected separately from a thread's live
connections, so tool availability alone does not describe the current runtime
state.
## What changed
- Add a nullable `runtimeStatus` to `mcpServerStatus/list` for thread-scoped
requests, covering not-started, starting, connected, authentication-required,
failed, cancelled, and disabled connections.
- Observe published connection state without starting or reconnecting servers,
and return an unknown status when no thread is supplied or the active
configuration no longer matches the published registration.
- Show connection state and tool counts in the compact `/mcp` view while
retaining the detailed inventory in `/mcp verbose` and compatibility with
servers that omit `runtimeStatus`.
## Testing
- Cover runtime status transitions, deferred and disabled servers, closed
transports, configuration changes, protocol compatibility, and TUI rendering.
GitOrigin-RevId: e3bb6efe652f0fa8b3c97d5c53e4729b3a87cd91
## What changed
- Scope executor-provided stop-hook calls to the MCP server environment that registered the hook, and reject calls when that environment does not match.
- Forward turn metadata with executor stop-hook requests while retaining the hook call's `threadId` metadata.
## Testing
- Cover environment matching for MCP tool calls and verify executor stop hooks receive their environment ID and request metadata.
GitOrigin-RevId: b544ea3a6ca1f61389fb4597978ed7aac27d529a
## What changed
- Add environment-provided MCP restrictions for configured and plugin-provided servers.
- Disable attachment-scoped servers while their environment configuration is pending, failed, or unselected, while preserving selected-plugin access and the controller-owned Apps server.
- Apply the resolved policy to runtime startup, model tool exposure, telemetry, OAuth flows, and skill dependency installation.
## Testing
- Cover policy filtering for configured and plugin servers, environment state transitions, Apps ownership, and managed OAuth rejection.
GitOrigin-RevId: 7ca5d4dda14068e758497f88835d5cba685e159e
## What changed
- Execute hook-triggered MCP calls through the runtime's latest connection set instead of preparing a separate catalog-bound call.
- Keep hook calls from waiting for server startup or reconnecting, and reject disconnected servers immediately.
- Apply the shorter of the hook-requested timeout and the server's configured tool timeout.
## Testing
- Update connection manager tests to cover non-waiting calls and disabled tool filters.
GitOrigin-RevId: 3166b2d1e26d9ae2b8b3a648b4d0897cc260eb93
## Why
MCP runtime reconciliation can occur while an unchanged server is still
starting, such as when capability roots become available. Reconciliation
should neither wait for that startup nor initialize the same server again.
## What changed
- Reuse an active, uncancelled pending MCP connection when its identity,
catalog limit, and protocol mode still match.
- Publish the shared startup result through the reconciled server view while
applying its updated tool filter.
- Preserve cancellation when a reused pending server is subsequently removed.
## Testing
Added coverage for filtered-tool publication, disabling a reused pending
server, and capability-root reconciliation during Codex Apps startup.
GitOrigin-RevId: 690cf8695e75b341d097d7d22b5da6993b24e787
## Why
When an MCP server rejects existing OAuth credentials, reporting that the server
is merely not logged in obscures that the user needs to authenticate again.
## What changed
Pass the MCP startup failure reason into error formatting and report that the
server requires OAuth reauthentication when appropriate. Preserve the existing
recovery hint for either `codex mcp login` or the client's OAuth sign-in flow.
## Testing
Cover the reauthentication message for both login flows and verify it in the
streamable HTTP OAuth round-trip test.
GitOrigin-RevId: 6f6d5fe7519b1a476f8b70d976223b2ea1ec938e
## What changed
- Add `http_headers_helper` configuration for local streamable HTTP MCP servers. The configured shell command runs once per connection and returns a JSON object of headers that is cached across requests.
- Apply helper headers to MCP startup and OAuth flows while restricting them to the server origin, stopping redirects, rejecting reserved or duplicate headers, and enforcing output and execution limits.
- Reject helpers for remote or managed-disabled servers, use the local environment working directory, and redact helper commands from `codex mcp list` and `codex mcp get` output.
## Testing
- Cover configuration validation, helper lifecycle and output parsing, origin isolation, OAuth discovery and token refresh, managed requirements, environment selection, and CLI redaction.
GitOrigin-RevId: 84e0e26ce75520b0869d37c72b1678e033bd6818
## Why
Subagents can reuse cached MCP tool definitions without opening another server
connection immediately, but marking a server as required previously forced eager
startup even when its cached tools were available.
## What changed
- Allow required MCP servers with usable cached tools to remain dormant under
the lazy startup policy.
- Treat those cached tools as satisfying required-server validation and tool
catalog capture until the server is needed.
- Preserve eager startup for required servers without cached tools and start a
dormant server when one of its tools is called.
## Testing
Extend the cached HTTP MCP subagent test to cover both optional and required
servers, verifying that the required server stays dormant until its cached tool
is called.
GitOrigin-RevId: 48ea20b9ae447eef27461caee2320ea78465bb6a
## Why
Subagents should be able to use known HTTP MCP tool definitions without opening a connection until a tool is actually called.
## What changed
- Add streamable HTTP configurations to the process-scoped MCP tool catalog cache when their authentication identity can be derived safely.
- Fingerprint transport settings, relevant environment variables, protocol mode, plugin status, and client capabilities so catalogs are reused only across equivalent connections.
- Keep OAuth and other dynamically resolved credential configurations out of the shared cache.
## Testing
- Add an integration test proving that a subagent receives cached HTTP tools without reinitializing the server, then starts its own connection when it calls one.
GitOrigin-RevId: bb2787e29e9b71993def31aafd1f0d8f1b728d26
## Why
Refreshing MCP connection identities should not stall the async executor while another process holds the OAuth credential-store lock.
## What changed
- Probe file and secrets credential stores without waiting during runtime refreshes, retaining the last known credentials when the store is contended.
- Reuse unchanged authentication failures while credentials are temporarily unavailable instead of restarting the connection.
- Run blocking credential loads on a blocking worker during MCP client construction.
- Allow OAuth fallback when a configured authorization header is missing, empty, or invalid.
GitOrigin-RevId: 38dad5606da9a1ceb30fe945000ec9151a5058fa
## What changed
- Add lifecycle-backed gauges for in-flight and queued app requests, pending server requests and mailbox messages, active turns, and live MCP connections.
- Register each gauge on first use and decrement it when the tracked work or resource is dropped.
- Document the new `server/diagnostics` gauge names and account for the diagnostics request itself in the response example and test.
GitOrigin-RevId: 4bb82a3918af0b502d149b5d761ad669b2083f39
## Why
An OAuth-backed Streamable HTTP MCP server can fail startup when its stored credentials are rejected. If a client then completes OAuth sign-in and replaces those credentials, the failed server needs to become available without restarting the Codex session.
## What changed
- Track the credential store and OAuth tokens used for each MCP connection, and detect replacements after authentication failures.
- Refresh affected MCP servers before the next turn and include their recovered tools in the model request.
- Classify HTTP authentication challenges and rejected refresh tokens as requiring reauthentication, with sign-in guidance appropriate to local and client-managed environments.
## Testing
- Cover recovery from expired credentials through an externally written replacement token.
- Cover authentication-error classification, startup messaging, and `Auto` store migration from file credentials to the keyring.
GitOrigin-RevId: 9b090089d4e7a44ffc182f4e25da20a18e97b70f
## Why
Subagents can reuse cached MCP tool definitions without starting every optional
server before any of its tools are used.
## What changed
- Defer startup for subagent-owned, optional MCP servers when their cache has at
least one visible, allowed tool, and trigger startup on the first operation
that needs the live client.
- Keep root sessions, required servers, selected plugin servers, and explicit
reconnects eager.
- Exclude empty catalogs from cache eligibility and keep dormant servers out of
the initial startup summary.
## Testing
- Extend MCP tool-cache coverage for eager root startup, unused and filtered
subagent servers, deferred startup status events, and startup that survives an
interrupted first tool call.
GitOrigin-RevId: e7528ff08312583a6747cbc6550ea5b6eabda9a6
## What changed
- Cache an `McpBinding` on each published MCP runtime once server startup has
settled and the tool catalog revision is stable.
- Reuse the cached binding while the catalog revision remains unchanged, and
rebuild it after a runtime publish or catalog change.
- Treat terminal failures from optional servers as stable, while continuing to
wait for required servers and reconnecting Codex Apps servers.
## Testing
- Cover binding reuse within one published runtime and invalidation after a
republish.
- Cover stability checks for pending, failed optional, required, and
reconnected servers.
GitOrigin-RevId: 783a6b04897f6c484a6d56fd9cee542814041759
## Why
App-server clients need to advertise structured MCP extension settings, including supported MCP App UI MIME types, rather than only opting into OpenAI form elicitation with a boolean.
## What changed
- Add an `extensions` map to initialize capabilities and preserve `mcpServerOpenaiFormElicitation` as a legacy alias for `openai/form`.
- Capture the declared extension profile when a thread is started, resumed, or forked, propagate it to subagents, and advertise it to downstream MCP servers during initialization.
- Keep the profile stable for the lifetime of the loaded session instead of changing it on later turns or direct tool calls.
## Testing
- Cover extension conversion, downstream MCP initialization, session isolation, legacy form support, and subagent inheritance.
GitOrigin-RevId: fbcedbb74ce788e574b0f884a4c45c4cedb9de54
## Why
Codex Apps tool catalogs can exceed the standard MCP limit of 2,048 items and
still need to remain available through tool search.
## What changed
- Allow up to 8,192 catalog items for host-owned `codex_apps` registrations.
- Keep the 2,048-item limit for other MCP servers, including extensions that
register the `codex_apps` name.
- Restrict the shared Codex Apps tool cache to host-owned registrations.
## Testing
- Cover large catalogs for built-in and hosted-plugin registrations while
verifying that another extension retains the standard limit.
- Verify that a same-named extension cannot read the host-owned Apps cache.
GitOrigin-RevId: 2265693a460e9be63c4c849169cba7324def4f49
## Why
MCP servers running in an executor must not reuse host-owned OAuth credentials,
even when they have the same configured server name and URL.
## What changed
- Derive environment-scoped credential names for executor-owned MCP servers and
use them consistently for login, logout, authentication, and connection setup.
- Mark executor-owned entries in the file credential store and fail closed when
loading or saving an entry could cross the host/executor boundary.
- Preserve compatibility with existing local OAuth credentials, including local
server names that overlap the new reserved prefixes.
## Testing
- Extend the executor MCP integration test to verify that executor requests use
the executor token, never send the host token, and persist both credentials
separately.
GitOrigin-RevId: 4fc92d0533b0fe2e0df34f6d47b81e3d20d07807
## Why
A pending optional MCP server can delay the first model request even when the
turn does not need that server.
## What changed
- Give optional MCP servers a shared one-second startup grace, then omit servers
that are still pending from the captured tool catalog.
- Continue waiting when the turn explicitly requires a server through a plugin,
skill dependency, or `mcp://` mention, and preserve this behavior for input
received between sampling steps.
- Route direct resource requests to the live connection set when a pending
server was omitted from the binding, while keeping all-server resource
discovery non-blocking.
## Testing
- Cover the shared grace period and resource behavior for pending optional
servers.
- Verify plugin, plugin-skill, and direct MCP mentions wait for startup.
- Verify an Apps-enabled turn proceeds without tools from an unrelated pending
optional server.
GitOrigin-RevId: b5f895c5a5362fe73f7d33250367662d4a217e4d
## What changed
- Add an opt-in `mcp_2026_07_28` protocol mode while preserving the legacy
lifecycle by default.
- Negotiate the new protocol over streamable HTTP with `server/discover`,
including bounded responses, redirect protection, and fallback only when a
response establishes that the endpoint is legacy-only.
- Require stdio servers to opt in with
`CODEX_MCP_PROTOCOL_VERSION=2026-07-28`, and add a bounded local stdio
transport for the modern lifecycle.
- Consume paginated tool, resource, and resource-template catalogs in modern
mode, reject repeated cursors, and retain discovered server identity.
- Reconnect reusable MCP clients when their selected protocol mode changes.
## Testing
- Cover HTTP JSON and SSE discovery, legacy fallback and rejection cases,
redirects, retries, response limits, and pagination.
- Cover local and executor stdio discovery, protocol markers, message limits,
and legacy compatibility.
GitOrigin-RevId: f6a78816e127d2a482292d63b91c8384f1595903
## Why
Cached MCP definitions can be supplied to inference without waiting for the
server to finish initializing.
## What changed
- Publish cached tools while startup is still in progress, clearing their
potentially stale read-only hint.
- Wait for the selected server to start before executing a tool call, then
prepare the call against the refreshed live binding.
- Keep cached tools visible in a binding even when no live client is available,
while rejecting attempts to prepare those calls.
## Testing
- Cover cached-tool visibility before startup and replacement with live tool
metadata afterward.
- Verify cached definitions reach inference before MCP initialization and that
calls unavailable in the live catalog return the expected model-visible
error.
GitOrigin-RevId: 3aae8f474c344ccdc5e08fe321bbad21d85bffd1
## What changed
- Accept a table form of `features.non_prefixed_mcp_tool_names` with an
optional `server_names` list while preserving the existing boolean form.
- Omit the legacy `mcp__` namespace prefix only for tools from selected MCP
servers. When no server list is provided, the enabled feature continues to
omit the prefix for every server.
- Cover configuration resolution, tool normalization, and an MCP stdio
round trip with selected servers.
GitOrigin-RevId: bdfb7ac54226de5051f06610e2c6b78b23912ef0
Keep the `McpConnectionSet` test constructors and accessors alongside the
connection manager tests instead of defining test-only helpers in the
production module.
GitOrigin-RevId: 925392f8b7d903b131ef1b49afd7e50f41226b1f
## Why
MCP connection reconciliation can otherwise reuse a client after its service or
underlying transport has stopped.
## What changed
- Add `RmcpClient::is_closed` to report closed client, service, and transport
states.
- Exclude closed clients when selecting a reusable MCP connection so
reconciliation starts a replacement.
## Testing
- Add a reconciliation test that disconnects an in-process MCP transport and
verifies that its connection is replaced.
GitOrigin-RevId: 366a85c9959dbe1763923d7e33a4ecf52d9c3c6a
## Why
Refreshing MCP runtime state should not restart an unchanged, ready server or
relist its tools.
## What changed
- Reconcile refreshed MCP configuration against the published connection set
and reuse connections whose transport, environment, authentication, OAuth
credentials, and client capabilities are unchanged.
- Keep tool filters, timeouts, metadata, and plugin provenance in the published
server view so those settings can change without reconnecting.
- Update elicitation authority in place for reused connections, and reconnect
when connection-defining inputs or live OAuth credentials change.
## Testing
Added coverage for unchanged-server reuse, view-only and elicitation-policy
updates, connection identity changes, OAuth credential changes, and avoiding
redundant tool listing.
GitOrigin-RevId: f04f5db5fbdde127b6a14f6aa9673112c0b557e7
## What changed
- Make `McpRuntime` own the published MCP configuration, connections, elicitation routing, and selected capability roots for a thread.
- Capture immutable MCP bindings for model steps and tool calls so in-flight work keeps a consistent connection set and approval authority while refreshed state is published atomically.
- Mark MCP state dirty when relevant configuration, plugins, environments, authentication, or elicitation capabilities change, then rebuild it before the next sampling request or out-of-band MCP operation.
- Separate config reloads from server invalidation: `ReloadMcpConfig` applies resolved MCP inputs, while `RefreshMcpServers` requests reinitialization from the thread's latest state.
## Testing
- Cover refreshed state visibility for existing turns, stable step bindings, current approval authority, cancelled refresh retries, resource-client reconciliation, and Apps recovery between sampling requests.
GitOrigin-RevId: 59eabb1aa8dc083426bd18ef4d3630508f376401
## What changed
- Use `McpConnectionSet` as the primary name for the collection of active MCP
server connections throughout `codex-mcp` and its callers.
- Keep `McpConnectionManager` as a backward-compatible type alias.
GitOrigin-RevId: 9753838b4fd0a0b6429a9c4af0c077137616f388
## Why
An MCP connection or tool catalog can change after a model step captures its
available tools. Calls from that step must not be rerouted to a replacement
client or run against a catalog revision the model did not see.
## What changed
- Add `McpBinding` to capture the ready clients, visible tools, resources, and
server metadata for one sampling request.
- Prepare calls against the captured client and reject them if the tool catalog
changes before call preparation and execution finish.
- Keep resource access and cache identity tied to the binding's exact connection
set, and omit servers that do not have a ready client.
- Preserve each client's unfiltered tool catalog so bindings can apply the
current filter consistently.
## Testing
Add coverage for connection replacement, stale catalog rejection, serialized
call preparation and refresh, ready-client capture, and binding-scoped resource
identity.
GitOrigin-RevId: a25da4d2c7f5ed06c41caa0c6d15c464c1698314
## What changed
- Add `McpBindingClients` to hold the ready MCP clients captured for a model step.
- Move concurrent, paginated resource and resource-template listing into the new helper.
- Keep server filtering in `McpConnectionManager` while preserving per-client timeouts, duplicate-cursor detection, and warning behavior.
GitOrigin-RevId: 0e3f8f135126b17d73058ca84e98493a27543ca0
## What changed
- Move required-server startup validation into `connection_manager/required.rs`.
- Move tool listing, lookup, metadata attachment, and Codex Apps cache refresh into `connection_manager/tool_catalog.rs`.
- Keep the existing connection-manager API and behavior unchanged.
GitOrigin-RevId: 7249896699f756bdbb6e00bc2bbc4a98098741a9
## What changed
- Make the MCP connection manager's event sender optional.
- Skip startup notifications when a caller has no session event stream.
- Decline interactive MCP elicitations in that mode while allowing the
underlying operation to continue.
- Update threadless resource reads, status snapshots, and connector discovery
to use the channel-free path.
## Testing
- Extend the app-server MCP resource test to verify that a threadless resource
read declines an elicitation and still returns the resource contents.
GitOrigin-RevId: 65fea152adbc02a78e09a40cae4134b6dccffd34
## Why
Starting a new session should not have to wait for an unchanged stdio MCP
server to initialize before its tools can be presented to the model.
## What changed
- Cache recent tool catalogs for matching stdio MCP server configurations and
use them while a new live connection starts in the background.
- Bound the process-wide cache to 32 entries with a 30-minute TTL, and key
entries by the server configuration, environment, working directory, and
elicitation capabilities that affect the catalog.
- Keep connection-specific instructions and tool annotations out of cached
catalogs. Resolve tool calls against the live connection so tools that are
unavailable or hidden in the new session cannot be invoked from stale data.
- Publish only the newest completed catalog fetch and do not cache HTTP servers
or stdio configurations that depend on remotely sourced environment values.
## Testing
- Add coverage for cache expiration, environment changes, catalog sanitization,
and out-of-order refreshes.
- Add an end-to-end test that starts inference from cached definitions while a
replacement MCP process initializes, then verifies calls and visibility
against the live process.
GitOrigin-RevId: 2ed0f29ba53ace12532730f951b175c588aa3ae7
## Why
OAuth metadata discovery for an optional HTTP MCP server can stall, delaying
`thread/start` even though that server is not required.
## What changed
- Remove eager MCP authentication-status discovery from session and connection
manager startup.
- When an MCP connection reports an authentication error, inspect configured and
stored credentials locally to preserve login and reauthentication failure
reasons without making another discovery request.
- Continue using server configuration directly for startup error messages and
timeout hints.
## Testing
Add an app-server regression test with an optional HTTP MCP endpoint that accepts
a connection but never responds, and verify that `thread/start` completes without
waiting for its configured startup timeout.
GitOrigin-RevId: 56fe0188cc1ceaf746b0b65e8b7dd1fd2e13420d
## What changed
- Apply each server's `startup_timeout_sec` (or the default) while creating the
MCP client, so the deadline also covers transport setup.
- Launch local stdio servers on a blocking task so synchronous command
resolution and process creation do not prevent the deadline from firing.
- Recognize the new client-startup timeout error and show the existing
`startup_timeout_sec` configuration hint.
## Testing
- Extend the timeout error display test to cover the client-startup timeout.
GitOrigin-RevId: 1967c62f943d55f6aa18792d4488e52c22f1e717
## What changed
- Move the Codex Apps tool cache into a reusable `codex-connectors` runtime manager keyed by account and workspace.
- Represent cached tools as atomically published snapshots with refresh timestamps while preserving the newest accepted fetch generation.
- Harden disk persistence with bounded reads, atomic file replacement, and serialized writes so an older fetch cannot overwrite newer state.
## Testing
Add coverage for identity isolation, snapshot timestamps, oversized cache files, atomic replacement, and concurrent persistence ordering.
GitOrigin-RevId: 5ea2234469daae3abf54b030244c3251de62ca5a
## Summary
Codex Apps file parameters are exposed to the model as local paths,
uploaded at execution time, and rewritten into provided-file payloads
before the MCP tool call.
The rewrite currently includes the documented optional fields
`mime_type` and `file_name` for every file parameter. Apps with strict
schemas can reject those fields when they are not declared.
## Changes
- Derive the supported optional file fields from each
`openai/fileParams` parameter's raw input schema before replacing it
with the model-visible local-path schema.
- Always include `download_url` and `file_id`.
- Include `mime_type` and `file_name` only when that specific file
parameter's schema accepts them, including schemas that allow additional
properties.
- Handle scalar and array file parameters, including items-only arrays,
composed schemas, and local JSON Schema references.
- Preserve the existing restriction that only the host-owned Codex Apps
MCP server can use this upload path.
This supports tools with different file contracts in the same app: one
parameter can accept the optional fields while another remains strict.
## Validation
- `just test -p codex-mcp`
- `just test -p codex-core mcp_openai_file`
- `just test -p codex-core codex_apps_file_params_`
- `just fix -p codex-mcp`
- `just fix -p codex-core`
- `just fmt`
- `git diff --check`
- Manually verified in the Codex Electron app that:
- a strict file schema receives only `download_url` and `file_id`
- a rich file schema also receives `mime_type` and `file_name`
Related: #31330
## Why
Every MCP tool-list build emitted two normal-path TRACE events per
configured server: one before waiting for tools and one after listing
them. On active sessions this produced thousands of nearly identical
SQLite rows while carrying little information beyond server readiness
and tool counts.
## What changed
- Remove the two normal-path per-server TRACE events.
- Keep the existing per-server trace span for timing and remote trace
context.
- Emit per-server details only when a server's tools are unavailable.
- Emit one bounded summary per tool-list build with available server,
unavailable server, and tool counts.
Successful builds retain the useful aggregate signal without repeating
it for every server.
Related to #28224.
[Codex Thread
019f2408-dc59-79f2-b245-4c11debd1a61](https://codex-thread-link.openai.chatgpt-team.site/thread/019f2408-dc59-79f2-b245-4c11debd1a61)
## Why
Long-lived Codex sessions can outlive the ChatGPT bearer token that was
present when the MCP runtime started.
The Responses path already recovers from token expiration by refreshing
or reloading the shared `AuthManager`. The reserved `codex_apps`
hosted-plugin client did not observe that update: `McpConnectionManager`
built its `/ps/mcp` HTTP auth once from a `CodexAuth` snapshot, and
`auth_provider_from_auth` copied that snapshot bearer into a static
`BearerAuthProvider`.
After the copied bearer expired, `/ps/mcp` kept sending it even though
Responses had a newer token in the same `AuthManager`. The failure
occurred before downstream connector execution, so unrelated apps such
as Gmail, Slack, and Google Calendar could all fail with the same
transport-level `401 token_expired`.
This replaces
[openai/codex#29474](https://github.com/openai/codex/pull/29474), which
was closed for inactivity without being merged. A new long-lived-session
report reproduced the same simultaneous `/ps/mcp` expiry pattern across
unrelated apps.
## What changed
- Add an `AuthManager`-backed request-header provider in
`codex-model-provider`. It keeps an `Arc<AuthManager>` and reads
`auth_cached()` for each outbound request, so the next `/ps/mcp` call
sees a token refreshed by the existing Responses/auth-recovery flow.
- Scope that provider to the startup account, ChatGPT user, and
workspace identity. Same-identity token reloads are followed; an account
switch emits no ambient auth until account-scoped MCP state is rebuilt.
- Have `McpConnectionManager` construct the dynamic provider only for
the reserved `codex_apps` registration used by the hosted-plugin
`/ps/mcp` path.
| MCP path | Auth behavior after this change |
| --- | --- |
| Reserved `codex_apps` hosted-plugin `/ps/mcp` | Read current
same-identity auth from the shared `AuthManager` per request |
| `codex_apps` with `CODEX_CONNECTORS_TOKEN` | Keep the environment
bearer-token override |
| User-configured/direct MCP registrations | Keep their existing
configured auth path |
## Non-goals
- No plugin-service changes.
- No downstream Slack, Gmail, Calendar, or other connector
OAuth/link-refresh changes.
- No auth UI changes.
- No behavior change for user-configured/direct MCP registrations.
- No new `/ps/mcp`-initiated token refresh; this makes `/ps/mcp` observe
refreshes already performed through the shared `AuthManager`.
## Tests
- `just test -p codex-model-provider`
- Covers same-identity token reloads and refuses a changed startup
identity.
- `just test -p codex-mcp`
- `just test -p codex-core mcp_auth_refresh`
- Creates the reserved hosted-plugin `codex_apps` `/ps/mcp` client
before the shared `AuthManager` changes, updates that same manager
through its public external-auth path, performs a real `tools/call`, and
asserts the request uses the current bearer.
## Summary
- measure successful legacy `app/list` latency with
`codex.apps.installed.duration_ms`, segmented by `path=legacy` and
`reload`
- measure successful host-owned `codex_apps` startup and explicit
refresh latency with `codex.apps.refresh.duration_ms`
- add the refresh trigger to successful
`codex.mcp.tools.fetch_uncached.duration_ms` samples for `codex_apps`
without changing other MCP-server samples
## Why
This establishes a small latency baseline for the current connector path
before `ConnectorRuntimeManager`, `app/installed`, and `app/read` land.
Error-rate and broader runtime-state instrumentation are intentionally
deferred.
This is telemetry-only and does not change connector behavior.
## Validation
- `just test -p codex-mcp` (94 passed)
- `just test -p codex-app-server list_apps` (13 passed)
- `just fix -p codex-mcp`
- `just fix -p codex-app-server`
- `just fmt`
- `git diff --check`
## Why
Code-mode tool results could return to the model while an MCP
elicitation was still waiting for user input. This differed from
parallel tool calling and could let the model continue before the user
resolved the request.
We need one session-level view of outstanding elicitations so tool
runtimes can consistently hold results until every pending elicitation
is resolved.
## What changed
- Added a counted, session-owned ElicitationService with RAII
registrations.
- Registered both core-originated and server-originated MCP elicitations
with the service.
- Migrated out-of-band elicitation tracking and unified exec timeout
pausing to the shared service.
- Made code-mode functions.exec and functions.wait capture their runtime
result normally, then hold it before returning while an elicitation is
outstanding.
- Kept terminate: true immediate; only its result is held.
- Preserved model-visible wall time across the elicitation hold.
- Kept the behavior session-scoped, with concurrent elicitations holding
the pause until all registrations are released.
## Problem
The built-in Codex Apps MCP client shares a future for the full startup
operation: connect, complete `initialize`, fetch the initial tools, and
return a usable client. Sharing deduplicates startup work, but it also
memoizes terminal errors.
After a transient connection, handshake, or initial `tools/list`
failure, later tool builds observe the same failed future. The thread
cannot reconnect after the backend recovers and continues serving its
startup-time cached tool snapshot, which may be empty or stale.
## Fix
When Apps MCP startup ends in an error, Codex starts bounded recovery
without putting startup latency on tool-router construction:
1. The current tool build immediately continues with the cached startup
snapshot.
2. After the initial failure is reported, Codex starts one fresh full
startup attempt in the background.
3. Concurrent tool builds share that in-flight attempt and also continue
with cached tools.
4. On success, the recovered client becomes active, refreshes the Apps
tools cache, emits a `Ready` startup status, and is reused by later
operations.
5. On failure, the cache remains unchanged and later tool builds may
start another background attempt after exponential cooldown: 1s, 2s, 4s,
8s, 16s, then 30s maximum.
Each recreated startup performs a fresh MCP `initialize` and uncached
`tools/list`. The MCP client retains its existing bounded retries for
retryable `initialize` and `tools/list` failures.
This avoids adding the Apps startup timeout to every request during a
sustained outage.
## Scope
This is limited to the built-in Codex Apps MCP client:
- no reconnects for user-configured MCP servers;
- no cache deletion; and
- no proactive refresh for a healthy client with stale tools.
## Tests
Coverage verifies:
- tool builds return cached tools without waiting for a blocked
reconnect;
- concurrent tool builds start only one background reconnect;
- failed reconnects preserve cached tools and respect exponential
cooldown;
- a recovered client is retained and reused; and
- a long-lived thread exposes recovered app tools on a later follow-up.
Validation:
- `just test -p codex-mcp` — 95 passed
- `just test -p codex-core
later_follow_up_uses_background_recovered_apps_after_mid_thread_startup_failures
--no-capture` — passed
- `just fix -p codex-mcp`
- `just fmt`
## Why
An MCP tool call can still be waiting for an elicitation response when
an environment update replaces the thread's MCP runtime.
Before this change:
```text
runtime A starts a tool call and asks the user
environment becomes ready, so runtime B is published
client answers the prompt through runtime B
runtime B cannot find runtime A's pending responder
```
The response is lost and the original tool call stays blocked.
## What changed
All MCP runtimes for one thread now share a small elicitation router:
```text
runtime A ---\
shared router: response token -> exact pending responder
runtime B ---/
```
When Codex surfaces an MCP elicitation, it assigns a unique opaque
response token. The router records which pending request owns that
token. A replacement runtime reuses the same router, so the latest
runtime can deliver a response to a request started by the previous
runtime.
The Codex-owned token also prevents two runtime connections that reuse
the same MCP server request ID from receiving each other's responses.
This does not retain or search old MCP managers. Only the pending
responder map is shared.
## Covered scenario
The integration test exercises the complete failure mode:
1. A thread starts while its selected environment is still unavailable.
2. A configured MCP server starts a tool call and asks the client for
input.
3. The environment becomes ready, causing Codex to publish a replacement
MCP runtime.
4. The client answers the original prompt after the replacement.
5. The original tool call receives that answer and completes.
A focused routing test also creates two runtimes with the same server
request ID and verifies that each response reaches the exact request
that emitted its token.
## Scope
This PR changes only elicitation response routing across MCP runtime
replacement. It does not change when runtimes are rebuilt, which
environments contribute MCP configuration, or how environment
availability is detected.
## Summary
- distinguish expired, non-refreshable stored MCP OAuth credentials from
first-time missing credentials
- carry a typed `failureReason: "reauthenticationRequired"` on the
existing `mcpServer/startupStatus/updated` notification only when user
action is required
- keep the public MCP auth-status API unchanged and regenerate the
app-server protocol schemas and documentation
## Why
An MCP server with an expired access token and no usable refresh token
currently fails startup without giving clients a reliable, typed
recovery signal.
The existing startup-status notification is the natural place to carry
this state. Its nullable `failureReason` keeps the recovery reason
attached to the failed startup transition without adding a one-off
notification. Internally, Codex distinguishes first-time login from
reauthentication and emits the reason only when the startup error itself
requires authentication.
## User impact
App clients can prompt an existing user to reconnect an MCP server when
automatic recovery is impossible by handling a failed
`mcpServer/startupStatus/updated` notification whose `failureReason` is
`reauthenticationRequired`. Starting, ready, cancelled, unrelated
failures, and first-time setup carry no reauthentication reason.
## Companion app PR
- openai/openai#1069582
## Validation
- `just test -p codex-app-server-protocol` — 248 passed; schema fixture
tests passed
- `cargo check -p codex-app-server -p codex-tui`
- `just test -p codex-rmcp-client -p codex-mcp` — 184 passed, 2 skipped
- `just test -p codex-protocol -p codex-app-server-protocol -p
codex-mcp` — 579 passed
- `just write-app-server-schema`
- `just fmt`
## Description
This makes Codex Apps tool reads use a shared in-memory snapshot instead
of rereading the disk cache every time `list_all_tools()` runs. Disk
still seeds the cache on startup and gets updated after successful
fetches, but it is no longer the live read path.
The core change is that `McpManager` now owns a process-scoped
`CodexAppsToolsCache`. Codex threads in the same app-server process now
share this Codex Apps in-memory tools snapshot. The snapshot is keyed by
the Codex home plus the Codex Apps identity: the active Codex auth
user/workspace and the effective Codex Apps MCP source config.
There's already code to hard-refresh the cache, so we respect it in this
PR.
## Local benchmark
I ran a local steady-state microbenchmark of the exact repeated Codex
Apps cached-tools read this PR removes, using the same real local cache
payload in both trees: `3,678,138` bytes and `381` tools. The cache file
was already warm in the OS page cache, so this measures same-process
reread/deserialization work rather than cold-disk latency or full turn
latency. Each run is 25 iterations (mimicking a turn that makes 25
inference calls).
| Version | Run 1 | Run 2 | Avg |
|---|---:|---:|---:|
| `origin/main` disk read + JSON deserialize + `filter_tools` | `50.755
ms` | `52.894 ms` | `51.825 ms` |
| This branch in-memory `current_tools` + `filter_tools` | `0.740 ms` |
`0.778 ms` | `0.759 ms` |
That removes about `51 ms` from each repeated Codex Apps cached-tools
read on this machine, roughly `68x` faster for that subpath. It is
useful evidence for the hot path this PR changes, but not a claim that
every production turn gets `51 ms` faster; end-to-end impact also
depends on the rest of `list_all_tools()` and tool-payload construction.
This is on my M2 Max macbook, so with a slower disk this would be much
worse (and indeed we did see this really blew up turn runtime with a
slow disk).
victor-openai