LibreChat/api/server/services/MCP.js
Danny Avila fa913148fb
🔒 fix: Refresh MCP OBO Tokens From the Live OpenID Session (#15334)
* 🧊 fix: Inline-refresh OpenID session tokens at MCP OBO call time

Resolves the walk-away failure mode where MCP tool calls using OBO auth
fail with "No valid OpenID access token is available for OBO exchange"
after a user idles past their access-token lifetime. The strategy-time
snapshot on `user.federatedTokens` could expire mid-stream before
`resolveOboToken` ran, while `req.session.openidTokens` carried a still-
valid (or refreshable) token that nothing read.

- New OpenIDSessionRefresh service: per-user single-flighted closure that
  reads `req.session.openidTokens` at OBO time and inline-refreshes via
  `openid-client.refreshTokenGrant` when expired (30s skew), persisting
  via `req.session.save()`. No cookie writes (headers already flushed).
- `resolveOboToken` gains a required UpstreamTokenProvider parameter
  (typed as `() => Promise<OIDCTokens | null>`, reusing the shared shape
  from @librechat/data-schemas). Compile-time guarantee that every call
  site is updated.
- New `session_refresh_failed` OboTokenResolutionReason distinguishes
  "session expired and IdP rejected refresh" from "no upstream token
  ever existed."
- `req` threaded through createMCPTool/createMCPTools/createToolInstance
  to construct the closure with captured request, plus fail-closed
  guards in MCPConnectionFactory.getOboTokens and MCPManager.callTool
  when the closure isn't plumbed.
- Startup warning in MCPServersInitializer when OBO is configured but
  OPENID_REUSE_TOKENS is unset (the strategy populating
  user.federatedTokens is only registered under reuse, so OBO would
  fail every call without it).

Tests: 16 new in OpenIDSessionRefresh.spec.js; obo.spec.ts extended
for the new param + error reason; wiring smoke tests in MCPManager,
MCPConnectionFactory, MCPServersInitializer, and MCP.spec.js.

* 🛡️ fix: Harden OBO inline-refresh against token type and session edge cases

- Token-preference asymmetry: live-token reuse and expires_at derivation
  now strictly gate on the access_token, not the id_token. Added a
  required `tokenPreference` parameter on isLiveSessionTokenStillValid,
  buildOIDCTokensFromSession, and createOpenIDSessionTokenProvider
  so every call site is explicit. Dropped the bogus id_token-exp
  fallback in performIdpRefresh — id_token TTL is governed by IdP
  session policy and would mark a short-lived access_token reusable
  past its real lifetime.
- Missing req in /reinitialize route: the manual reconnect
  endpoint now forwards req into reinitMCPServer, so OBO servers can
  build a session-aware upstream-token closure instead of failing with
  missing_upstream_token.
- Single-flight key collisions: composed key as
  tenantId:openidIssuer:openidId:sessionId via getSingleFlightKey.
  Concurrent calls in the same session still coalesce; separate sessions
  never share an in-flight refresh, preventing refresh-token rotation
  from breaking sibling sessions and preventing cross-tenant token
  crossover when distinct users share an IdP sub.
- Opaque access token reuse): persist accessTokenExpiresAt
  (unix seconds, from tokenset.expires_in) on each refresh AND on initial
  login / SPA refresh in setOpenIDAuthTokens. New getAccessTokenExp
  helper falls back to it when the access token isn't a JWT, avoiding
  redundant inline refreshes for Microsoft Graph and Auth0 default
  audiences.
- Log hygiene: the single-flight key (containing sessionId,
  openidId, openidIssuer, tenantId) is now SHA-256-hashed in the
  "Joining in-flight refresh" debug log. Preserves cross-line correlation
  via a 12-char prefix without leaking credential or PII material.

Documented req.session.openidTokens shape contract via JSDoc typedef so
the new accessTokenExpiresAt field has a discoverable home alongside the
existing accessToken/idToken/refreshToken/expiresAt/lastRefreshedAt.

Tests: OpenIDSessionRefresh.spec.js up to 30 passing (added coverage for
opaque-token reuse, JWT-access-token-exp fallback, no-id_token-fallback
regression, cross-session no-coalesce, persistence on refresh, and a
guard against stale accessTokenExpiresAt carryover). AuthService.spec.js
adds two cases covering accessTokenExpiresAt persistence on login.
mcp.spec.js (route) gains a regression test asserting req flows into
reinitMCPServer.

* 🔍 fix: Detect OBO-only MCP admin config overrides

Admin Config overlays for YAML-defined MCP servers compare only
ADMIN_CONFIGURABLE_FIELDS to decide whether to lazy-init a config-tier override.
The OBO config field was added after that fingerprint list, so an override that
only added or changed `obo` was treated as unchanged YAML and skipped.

Include `obo` in the admin-configurable field list and add a regression test for
an OBO-only override.

* 🔊 fix: Mock MCP OAuth timeout in SDK integration test

MCPConnectionFactory.attemptToConnect reads mcpConfig.OAUTH_HANDLING_TIMEOUT
when building the OAuth connection timeout. The SDK OAuth integration test
mocked mcpConfig without that field, which made the timeout calculation produce
NaN and caused the test to fail before the OAuth refresh/start path completed.

Add OAUTH_HANDLING_TIMEOUT to the test mock.

* ♻️ refactor: Pass OBO upstream-token closure into MCP instead of req

Build the OpenID upstream-token provider at the request boundary and thread
only the closure through MCP handling, so the MCP service layer no longer
receives the raw Express request. The closure still reads/refreshes the live
session at tool-call time, preserving the walk-away recovery.

- Drop `req`/`capturedReq` from createMCPTools, createMCPTool, reconnectServer,
  createToolInstance, and reinitMCPServer; forward `upstreamTokenProvider`
  instead. Closure is constructed in loadTools, loadToolDefinitionsWrapper, and
  the reinitialize route, where req/res are in scope.
- OBO: fall back to user.federatedTokens when the provider yields no live
  session, so OIDC remote-agent calls (verified bearer, no session) still work.
- Inline refresh: mirror a rotated refresh token to the refreshToken cookie via
  a shared setRefreshTokenCookie helper, guarded by !res.headersSent (no-op on
  the streaming path; session copy stays authoritative).
- Single-flight: hydrate a joining request's own session from the resolved
  tokens so a later OBO call doesn't replay a rotated-away refresh token.

Addresses owner feedback and three review findings.

* 🔒 fix: Recover OIDC refresh-token rotation after SSE OBO refresh

When an inline OBO refresh rotates the OpenID refresh token after SSE headers
have already been sent, the browser refreshToken cookie cannot be updated. Store
a short-lived encrypted bridge from the stale cookie token to the rotated token
so /api/auth/refresh can recover after express-session loss.

Use the signed openid_user_id cookie to load user context for bridge validation,
retry only on invalid_grant, and delete the bridge only after the bridged refresh
succeeds.

* 🔨 fix: hydrate joined OIDC refresh sessions with stable refresh tokens

Update single-flight OIDC refresh joiners whenever refreshed access token
state changes, even if the IdP keeps the refresh token unchanged.

This prevents joined requests from retaining stale accessToken or
accessTokenExpiresAt values and redundantly refreshing later in the same run.

* 🌉 Persist OIDC refresh-token recovery bridges in MongoDB

Store SSE OBO refresh-token recovery bridges in MongoDB instead of
process-local memory so /api/auth/refresh can recover after worker
restarts or cross-worker routing.

Derive bridge expiry from REFRESH_TOKEN_EXPIRY so the recovery window
matches the stale refreshToken cookie it repairs, and delete bridges
after successful recovery.

* 🤝 Coordinate OIDC inline refreshes across workers

Add a short-lived Mongo-backed refresh-flight record so concurrent
OBO refreshes for the same OpenID session do not redeem the same
rotating refresh token on different workers.

The winning worker performs the IdP refresh and stores an encrypted
result; joiners wait for that result, hydrate their request session,
and return without calling the IdP.

*  Keep OpenID marker cookies aligned on inline refresh

Refresh token_provider and openid_user_id with the same expiry as the
rotated refreshToken cookie when an inline OBO refresh can still write
headers.

Share the marker-cookie writer with the normal OpenID auth refresh path
so the fallback /api/auth/refresh branch continues to recognize valid
OpenID refresh tokens after session expiry.

* 🔑 fix: include refresh token in OIDC local refresh flight key

Key the process-local OIDC refresh coalescing by the current session
refresh token, matching the Mongo-backed flight key. This prevents a
request with a newly rotated token from joining an older pending refresh
and inheriting its failure/result.

* 🌉 fix: store OIDC refresh bridge without cookie response

Treat missing or non-cookie responses like headers-sent streaming
responses during inline OIDC refresh. When the IdP rotates the refresh
token and cookies cannot be written, persist a recovery bridge so a later
/auth/refresh can recover after session expiry.

* 🫙 fix: preserve stale OIDC cookie bridge key

Track the refresh token last written to the browser cookie separately
from the current session refresh token. When inline OIDC refreshes rotate
tokens without a writable response, keep bridging from the browser-stale
token directly to the latest session token.

* 🙌 fix: keep OIDC bridge recovery success on cleanup failure

Make refresh-token bridge cleanup best-effort after a bridged OIDC
refresh succeeds. A transient delete failure now logs a warning but does
not convert the already-refreshed session and cookies into a 403 response.

* 📦 test: Exclude RefreshTokenBridge from tenant-isolation coverage

Add RefreshTokenBridge to the tenant-isolation coverage allowlist because
refresh bridge lookups run during unauthenticated OpenID refresh recovery.
The controller first recovers user context from the signed OpenID marker
cookie, then the bridge methods apply explicit user and tenant filters.

Ambient tenant isolation would bind this recovery path to request-local
tenant context that is not available at the point the stale cookie is being
resolved

*  Fix OpenID refresh flight retry and marker hydration

Allow failed OpenID refresh flights to be reclaimed immediately instead of pinning transient errors.

Preserve the browser refresh-token marker when joined refreshes hydrate session tokens from a shared flight result.

Stabilize AuthService tests by isolating mocked module imports from prior suites.

* 🛠️ fix: centralize OBO identity scoping

Add shared auth identity helpers for app user ids, OpenID subjects,
tenant ids, and normalized OpenID issuers.

Thread a non-placeholder-visible OBO identity context from the real
request user through MCP connection, tool-call, reinit, and refresh
paths. Keep tenantId and openidIssuer out of createSafeUser so MCP
user placeholders do not expose those fields.

Scope OBO token cache and in-flight exchange keys by tenant, issuer,
OpenID subject, scopes, and a SHA-256 hash of the upstream assertion.
This prevents cross-tenant/cross-issuer collisions and avoids reusing
tokens minted from stale rotated assertions.

Use the shared identity helpers for OpenID refresh-flight keys and
refresh-token bridge recovery records so related OBO refresh paths share
the same identity normalization rules.

The helper is intended for auth-boundary and credential-cache code, not
as a blanket replacement for ordinary app user id ownership checks.

* 🛠️ fix: preserve OIDC refresh-token sync on save failures

Sync OpenID refresh-token cookie/bridge state before persisting the
session so a transient session-store failure cannot lose an IdP-rotated
refresh token.

Also trigger sync when the session refresh token differs from the
browser refresh-token marker, not only when the current grant rotates
the token. This lets later writable refreshes repair stale browser
cookies left behind by SSE refreshes.

Route refresh bridge identity through the shared identity helper with
the threaded OBO identity context, falling back to request/user context
when needed.

Add regression coverage for session-save failures, stale browser cookie
repair, non-writable bridge storage, and shared-helper identity fallback.

* 🛠️ fix: keep OIDC refresh bridge during recovery grace

After successful bridged refresh recovery, re-store the stale-cookie bridge
with a short grace TTL instead of deleting it immediately. This lets parallel
/api/auth/refresh requests that already sent the stale browser cookie recover
before they can observe the first response's Set-Cookie.

Retarget the bridge to the refresh token returned by the bridged retry so
B-to-C refresh-token rotation remains recoverable. The grace TTL is parsed with
math() and defaults to 60s, which shrinks the replay window from the original
REFRESH_TOKEN_EXPIRY bridge lifetime to the short recovery grace period.

Remove the now-unused explicit bridge delete path from the service and
data-schemas method surface. Add regression coverage for grace re-store,
identity symmetry, retry failure behavior, and same-key upsert replacement.

* 🛠️ fix: fail closed on OBO MCP user identity mismatch

Add an OBO-specific guard before MCP tool execution that requires the
effective invocation user and captured request user to both have ids and
to match. This prevents OBO tool calls from falling back to a separate
configurable.user_id identity after request-bound OBO context has already
been captured.

Keep the existing user id fallback behavior for non-OBO MCP calls.

Tests cover mismatched OBO users, missing user ids, and the matching-user
path ignoring a conflicting configurable.user_id.

* 🛠️ fix: Guard OpenID bridge retry user identity

Extract the shared OpenID refresh/user-resolution flow in AuthController
so the normal refresh path and bridge-recovery retry use the same grant,
claims, issuer, user lookup, and diagnostic logging code.

Preserve the existing path-specific behavior: the normal path still owns
migration updates and 401 login redirects, while the bridge retry still
falls through to the existing 403 invalid-token response.

Add a bridge-recovery guard that rejects retry results whose resolved
user id differs from the signed openid_user_id cookie before issuing
tokens or re-storing the grace bridge. Cover both the successful
matching-user recovery and the mismatched-user rejection.

* 🛠️ fix: type-safety polish on OBO data layer

Replace refresh token bridge query/update Record<string, unknown> usage
with typed Mongoose FilterQuery and UpdateQuery definitions.

Harden OpenID marker cookie JWT expiry handling by converting refresh
expiry milliseconds to integer seconds and rejecting invalid or
non-positive durations.

Add focused CSRF tests for fractional refresh expiry values and invalid
expiry configuration.

* 🛠️ fix: Bind OpenID session tokens to authenticated identity

Stamp OpenID session token state with the LibreChat user id, OpenID subject,
tenant id, and normalized issuer when tokens are stored.

Fail closed before OBO inline token reuse/refresh when the session token
identity does not match the current authenticated identity, preventing a stale
or mixed Express session from supplying another user's upstream assertion.

Also validate the normal /api/auth/refresh session-token reuse shortcut against
the signed marker-cookie user before returning cached session tokens.

Note: sessions created before this change carry no identity stamp and are
treated as a mismatch. This is self-healing — the reuse path forces a full IdP
refresh (which re-stamps the session) and the OBO path throws, surfacing as a
one-time re-authentication for active OBO users at deploy time. The session
re-stamps within one session lifetime (SESSION_EXPIRY, default 15 min).

* 🛠️ fix: Recover OpenID refresh token drift

Prefer the browser refresh-token cookie when it differs from the
server-side OpenID session state, and force a real IdP refresh in that
case instead of reusing stale session tokens.

Store a short-lived refresh-token bridge when inline OBO refresh writes
a rotated browser cookie but session persistence fails, so follow-up
refreshes can still recover from the old token.

Keep the bridge grace TTL centralized in RefreshTokenBridge so both
recovery paths use the same env-backed value.

Note: drift is measured against the last-synced browserRefreshToken
marker, so the SSE path (intentionally stale cookie, authoritative
session) does not false-positive. Sessions predating the marker have no
browserRefreshToken; for those, drift falls back to comparing the cookie
against the session refresh token and prefers the cookie on difference.
This is the same self-healing pre-change-session window as the identity
binding fix and re-syncs within one session lifetime.

Tests cover cookie/session drift selection, reusable-session bypass on
drift, bridge storage after session-save failure, and the shared bridge
constant wiring.

* 🛠️ fix: Harden OBO token caching and expiry handling

Reject malformed OBO grant responses before writing them to the exchanged-token cache so a missing access_token cannot poison the cache.

Store absolute expires_at values with cached OBO tokens and ignore legacy cache entries without usable expiry metadata. This keeps cached-token freshness based on the token’s real
remaining lifetime instead of reusing the original relative expires_in on cache hits.

Move OBO expiry normalization and skew helpers into packages/api and use them from both the JS exchange service and the TS MCP resolver. Apply a 30-second safety margin with a one-
second floor for short-lived tokens, covered by direct helper tests and caller-level regression tests.

Tests:
- packages/api: npm run build
- packages/api: npx jest src/mcp/oauth/expiry.spec.ts src/mcp/oauth/obo.spec.ts
- api: npx jest server/services/OboTokenService.spec.js

* 🛠️ fix: Harden OBO refresh-token bridge lookup and indexing

Reuse getValidOpenIDReuseUserId for the bridge-recovery user lookup in
refreshController instead of re-verifying openid_user_id inline. The shared
helper enforces the JWT_REFRESH_SECRET presence check and a strict
typeof payload.id === 'string' guard, rejecting tokens whose id claim is
present but not a string (e.g. a numeric id) that the inline check accepted.

Fail closed on issuer mismatch in getRefreshTokenBridge. Both the stored and
the expected issuer are now normalized and compared for equality, so a bridge
is recovered only when both sides agree (both absent, or both present and
equal after normalization). Previously the check was skipped whenever the
stored issuer was absent, allowing recovery across mismatched issuer context.

Drop the unused {oldRefreshTokenHash, userId, tenantId, openidIssuer} index
and the openidIssuer field on RefreshTokenBridgeQuery. The data-layer filter
only queries the 3-field {oldRefreshTokenHash, userId, tenantId} index; the
issuer is verified in application code, not the query. Hoist the repeated
model accessor into getRefreshTokenBridgeModel.

Note: issuer is now load-bearing for recovery. A bridge stored with an issuer
recovers only when the lookup supplies a matching issuer; the recovery lookup
reads user.openidIssuer via AUTH_REFRESH_USER_PROJECTION (an exclusion
projection that retains the field). If a user's persisted openidIssuer is
empty while the stored bridge has one, recovery fails closed (falls through to
normal re-authentication) until the bridge TTLs out — no security regression.

Tests cover invalid signed-cookie payloads bypassing the bridge, both
asymmetric issuer-presence cases, issuer normalization before comparison, and
an index-alignment assertion guarding against re-adding the dropped index.

* 🛠️ fix: Degrade OBO discovery on token resolution failures

Catch expected OboTokenResolutionError failures during MCP tool discovery and
fall back to unauthenticated tool listing instead of aborting discovery. This
keeps discovery aligned with the existing unauthenticated listing behavior while
preserving unexpected errors as real failures.

Also correct OBO tool-call freshness comment and tighten the OBO trust-check
permissions type to the existing role permission shape.

Tests:
- npx jest src/mcp/__tests__/MCPConnectionFactory.test.ts --runInBand --coverage=false
- npx jest src/mcp/oauth/obo.spec.ts --runInBand --coverage=false

* 🛠️ fix: tighten OBO tool-call errors, bridge logging, and flight typing

Move resolveToolCallUserId inside the tool-call try/catch so an OBO
identity mismatch surfaces with serverName/toolName context and the
standard tool-call-failed message instead of an opaque bare Error.

Raise the refresh-token bridge lookup failure log from debug to warn so
transient infrastructure failures on the unauthenticated /api/auth/refresh
path are observable, and guard the message access against non-Error values.

Replace the unknown+cast in isDuplicateKeyError with a hasErrorCode type
predicate so the duplicate-key check reads error.code without an assertion.

Preserve real math/isEnabled in the MCPConnectionFactory test mock (mock
only processMCPEnv) so mcpConfig timeouts no longer resolve to NaN, fixing
the TimeoutNaNWarning that masked slow OAuth retry behavior.

* 🧪 fix: Restore the Flight Uniqueness Index and Buffer the Graph Cache TTL

Two CI failures on the merge, both in suites this environment cannot run
(their MongoDB binary download is blocked).

`GraphApiService.spec.js` still asserted the unbuffered TTL. Graph tokens
route through the same `getTokenCacheTtlMs` as the OBO and openidStrategy
caches, so the entry now expires 30s before the credential does.

`openidRefreshFlight.spec.ts` dropped the database between tests, which takes
the indexes with it, and Mongoose builds them only once when the model is
compiled. Whether the unique `key` index survived into a test was a race with
that one-time build. Without it a second `create` inserts instead of raising a
duplicate-key error, so every worker believes it won the flight — the
mutual exclusion the file exists to prove. Indexes are now rebuilt after each
drop, which also makes the reclaim and complete cases reach those paths for
the right reason.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* fix: address OBO review findings

* 🔐 fix: Install Bridge Indexes and Carry OBO Through Assistant Recovery

Two findings from the Codex pass on d08c82f0d.

The refresh-token bridge relied on Mongoose auto-indexing for both of its
indexes, and `MONGO_AUTO_INDEX=false` is a supported deployment setting. A
bridge holds an encrypted refresh token and the TTL index is the only thing
that ever deletes one, so under that setting they would accumulate for the
life of the collection while concurrent upserts lost the compound uniqueness
the filter assumes. Installed before the first write, matching the flight
methods and the session and schedule methods before them.

`recoverServerTools`, the assistant create/update path that reruns
`reinitMCPServer` when a referenced server's catalog and connection snapshot
are both missing, was the last reinit site not carrying the upstream-token
closure. For an OBO server the factory rejects the connection outright, so the
assistant write failed with unavailable MCP definitions. It now builds the
provider at that request boundary like the other entry points; assistant
writes have no `res`, so a rotation there falls back to the recovery bridge.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* fix: harden OBO refresh coordination

*  fix: Keep Elapsed Expiries Elapsed and Revoke the Superseded Session

Two of the five findings from the Codex pass on fcdc15885 — the two that are
defects in code this branch introduced rather than design questions about the
bridge.

`getSkewedTokenExpiresAtMs` floored every result at a second in the future,
including an expiry the provider had already declared elapsed. An exchange
answering `expires_in: 0` or a past `expires_at` was handed to the MCP
connection stamped valid for another second, which only moves the failure
downstream. The floor now applies to a lifetime that is still live, which is
what it was for; an elapsed one stays elapsed so the caller rejects it. Same
for the cache TTL, which falls back to the elapsed-credential floor.

Bridge recovery left the stale token's durable Session behind. Only the token
it recovered through was passed as `existingRefreshToken`, so that one's
session was replaced while the token the browser actually presented kept its
record until its original expiry. That record, with the marker cookie still
bound to it, is what authorizes local image access for OpenID users — so a
copy of the stale cookie outlived the rotation it had lost. Revoked
explicitly on successful recovery.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* fix: close OBO refresh review findings

* 🎟️ fix: Carry the Bridged Token Through a Non-Rotating Recovery

Bridge recovery passes the browser's stale token as `existingRefreshToken` so
the durable Session naming it is the record replaced. That also makes the
stale token the fallback the installed session and the refresh cookie use when
a tokenset carries no `refresh_token` of its own, which holds only while the
recovery grant rotates.

An IdP that answers that grant without rotating sends the browser back to the
very token the bridge exists to retire: `storeOpenIDSession` installs and
deletes the same stale record in one call, and the cookie is rewritten to a
token the IdP already rejected — a sign-out on the next refresh. The grace
bridge one line above already guards this with `|| bridgedRefreshToken`; the
resolved tokenset now does the same, so leader and followers alike publish the
recovered token.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* 🚫 fix: Reject an OBO Exchange That Returns an Expired Credential

Preserving an elapsed expiry through the skew helper only helps if something
acts on it, and nothing did: `MCPManager.callTool` checks the access token and
nothing else before setting the Authorization header, so a credential the IdP
declared spent still went downstream to fail there. It is rejected at the
exchange now, where the reason is known, and retryably — the exchange itself
worked, so a fresh grant can succeed.

Completes the elapsed-expiry change in cb26a6f7d, which made the stamp honest
without giving anyone a reason to look at it.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* fix: close OpenID refresh review findings

* fix: coordinate OpenID refresh entry points

* chore: sort OpenID flight imports

* fix: fence OpenID refreshes during logout

* fix: close OpenID logout publication races

* fix: narrow completed refresh flight

* test: cover bridge cleanup failure after ownership loss

The compensating delete in storeRefreshTokenBridgeWithLease swallows its own
failure so the lease error stays the one the caller sees. Nothing asserted
that, so removing the inner catch left every suite green while callers began
receiving the cleanup error instead.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* fix: compensate OpenID bridges only on proven ownership loss

The post-write lease assertion deletes the bridge it just published when it
throws, but it threw for two different reasons: a coordination record that is
no longer ours, and a coordination read that simply failed. Treating the second
as the first destroys the only mapping from the token the browser still holds
to the one the IdP already rotated to, so a transient Mongo error on the
headers-already-sent path signed the user out.

Tag the ownership error where the lease raises it and compensate only for that,
preserving the bridge whenever ownership is merely undetermined. A preserved
bridge stays behind the logout revocation fence, so the safe default costs
nothing.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* fix: reject spent OpenID refresh results

Two ways a refresh could report success while handing back a credential
nothing can use.

An inline refresh carries the previous id_token forward when the IdP omits
one on rotation, so tokenset.id_token is not necessarily freshly issued.
setOpenIDAuthTokens applied its freshness guard only to the session copy and
took tokenset.id_token unconditionally, so /refresh returned an expired
bearer even though the grant produced a usable access token. Skip it only
when it is provably expired: an id_token whose expiry cannot be read stays
preferred, since access_token may be opaque or scoped to another audience.

normalizeExpiresIn preserves a zero or negative lifetime rather than
discarding it, so a grant declaring an already-spent access token still
published, rotating the refresh token and returning a token every freshness
check rejects. Each OBO call then repeated the grant. Reject an elapsed
lifetime before publishing; an unknown lifetime still publishes.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* fix: harden OpenID refresh publication

* fix: keep identity and results intact through OpenID refresh cleanup

Two follow-ons from the last round's fixes.

Stripping an expired carried-forward id_token from the refresh result removed
the only identity material a rotation without id_token leaves behind. The
result is rebuilt by buildOIDCTokensFromSession, so it carries no provider
claims() either, and getTokenClaims accepts only those two — bridge recovery
failed with "no usable identity claims" before setOpenIDAuthTokens could hand
back the fresh access token. The stripped token now travels in a
non-enumerable marker, alongside the existing browser and predecessor markers,
which identity resolution reads and the authentication response never sees.

The lease drained a pending renewal by awaiting it in finally, so a transient
coordination failure there threw from finally and replaced the operation's
result. The refresh had already settled and published, so the caller saw a
failure on credentials that had rotated. Proven ownership loss is recorded on
ownershipLost and checked before the return, so the drain has nothing to add
but noise; it now absorbs and logs.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SxKWxwqxAGckYpRsYTqx3F

* refactor: fence OpenID recovery publication

* fix: close OpenID publication transaction

* refactor: make OpenID publication transactional

* fix: satisfy OpenID publication type checks

* fix: fence OpenID session publication

* fix: authorize OpenID refresh publication

* fix: bind OpenID replay generations

* fix: fence OpenID response generations

* fix: authorize OpenID token delivery

* fix: linearize OpenID publication delivery

* chore: sort OpenID refresh flight imports

---------

Co-authored-by: J.C. Bartle <jcbartle@users.noreply.github.com>
Co-authored-by: jbartle <jbartle@rand.org>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: jcbartle <7274202+jcbartle@users.noreply.github.com>
2026-08-29 23:55:58 -04:00

1654 lines
60 KiB
JavaScript

const { tool } = require('@librechat/agents/langchain/tools');
const { logger, getTenantId } = require('@librechat/data-schemas');
const { Providers, Constants: AgentConstants } = require('@librechat/agents');
const {
sendEvent,
PENDING_STALE_MS,
MCPOAuthHandler,
MCPTokenStorage,
isMCPDomainAllowed,
splitMCPToolKey,
normalizeServerName,
normalizeMCPToolKey,
stripServerNamePrefix,
stripServerNamePrefixes,
buildServerNameAliases,
findShadowedServerNames,
getAssistantToolDefinitions: loadAssistantToolDefinitions,
toProviderToolDefinition,
resolveMCPServerContext,
normalizeJsonSchema,
GenerationJobManager,
resolveJsonSchemaRefs,
sanitizeGeminiSchema,
buildMCPAuthStepId,
buildMCPAuthToolCall,
processMCPEnv,
preProcessGraphTokens,
buildMCPAuthRunStepEvent,
buildMCPAuthRunStepDeltaEvent,
buildMCPAuthRunStepEndDeltaEvent,
isUserSourced,
hasCustomUserVars,
checkAccessWithRequestCache,
getMissingCustomUserVars,
getUserMCPAuthMap,
getServerCustomUserVars,
requiresEphemeralUserConnection,
requiresOAuthMachinery,
hasRuntimeUrlPlaceholders,
containsGraphTokenPlaceholder,
createAuthIdentityContext,
isOAuthServer,
OpenIDReauthRequiredError,
} = require('@librechat/api');
const {
Time,
CacheKeys,
Constants,
Permissions,
PermissionTypes,
isAssistantsEndpoint,
} = require('librechat-data-provider');
const {
getOAuthReconnectionManager,
getMCPServersRegistry,
getFlowStateManager,
getMCPManager,
} = require('~/config');
const db = require('~/models');
const { findToken, createToken, updateToken, deleteTokens, findPluginAuthsByKeys } = db;
const { getGraphApiToken } = require('./GraphTokenService');
const { exchangeOboToken } = require('./OboTokenService');
const { createOboTrustChecker } = require('./OboPolicyService');
const { createOpenIDSessionTokenProvider } = require('./OpenIDSessionRefresh');
const { reinitMCPServer } = require('./Tools/mcp');
const {
getAppConfig,
getCachedTools,
getMCPServerTools,
cacheMCPServerTools,
} = require('./Config');
const { getLogStores } = require('~/cache');
const MAX_CACHE_SIZE = 1000;
const lastReconnectAttempts = new Map();
const RECONNECT_THROTTLE_MS = 10_000;
const missingToolCache = new Map();
const MISSING_TOOL_TTL_MS = 10_000;
async function userCanUseMCPServers(user, req) {
if (!user?.id || !user?.role) {
return false;
}
try {
return await checkAccessWithRequestCache({
req,
user,
permissionType: PermissionTypes.MCP_SERVERS,
permissions: [Permissions.USE],
getRoleByName: db.getRoleByName,
});
} catch {
logger.error(`[MCP][User: ${user.id}] Failed MCP permission check`);
return false;
}
}
function createMCPPermissionContext(req) {
return {
canUseServers: (user = req?.user) => userCanUseMCPServers(user, req),
};
}
function evictStale(map, ttl) {
if (map.size <= MAX_CACHE_SIZE) {
return;
}
const now = Date.now();
for (const [key, timestamp] of map) {
if (now - timestamp >= ttl) {
map.delete(key);
}
if (map.size <= MAX_CACHE_SIZE) {
return;
}
}
}
const unavailableMsg =
"This tool's MCP server is temporarily unavailable. Please try again shortly.";
function getOAuthFlowId(userId, serverName, tenantId = getTenantId()) {
if (!tenantId) {
return MCPOAuthHandler.generateFlowId(userId, serverName);
}
return MCPOAuthHandler.generateFlowId(userId, serverName, tenantId);
}
async function getAppConfigForRequest(req) {
const user = req?.user;
return await getAppConfigForUser(user?.id, user);
}
async function getAppConfigForUser(userId, user) {
return await getAppConfig({ role: user?.role, tenantId: getTenantId(), userId });
}
/**
* Resolves config-source MCP servers from admin Config overrides for the current
* request context. Returns the parsed configs keyed by server name.
* @param {import('express').Request} req - Express request with user context
* @returns {Promise<Record<string, import('@librechat/api').ParsedServerConfig>>}
*/
async function resolveConfigServers(req) {
try {
const registry = getMCPServersRegistry();
const appConfig = await getAppConfigForRequest(req);
return await registry.ensureConfigServers(appConfig?.mcpConfig || {});
} catch {
logger.warn('[resolveConfigServers] Failed to resolve config servers; degrading to empty');
return {};
}
}
/**
* Resolves operator-managed MCP server names from admin Config overrides for the current request.
* Returns a request-time snapshot for DB server creation, not a cross-process lock.
* @throws Propagates app config lookup errors to keep DB server creation fail-closed.
* @param {import('express').Request} req - Express request with user context
* @returns {Promise<string[]>}
*/
async function resolveMcpConfigNames(req) {
const appConfig = await getAppConfigForRequest(req);
return Object.keys(appConfig?.mcpConfig || {});
}
/**
* All configured server names in the normalized form tool keys are built with.
* Unlike `resolveConfigServers`, this keeps unmodified YAML servers, which
* `ensureConfigServers` skips - those are exactly the ones that must still
* resolve the tool-key boundary.
* @param {import('express').Request} req
* @returns {Promise<string[]>}
*/
async function resolveMcpServerNames(req) {
try {
const names = await resolveMcpConfigNames(req);
return names.map(normalizeServerName);
} catch (error) {
logger.warn(
'[resolveMcpServerNames] Failed to resolve server names, degrading to empty:',
error,
);
return [];
}
}
/**
* Config-source servers and all configured names from a single app-config read,
* so the tool-loading path does not pay two lookups for the same principal.
* Degrades to empty like `resolveConfigServers` rather than aborting tool loading.
* @param {import('express').Request} req
* @returns {Promise<{ configServers: Record<string, import('@librechat/api').ParsedServerConfig>, serverNames: string[] }>}
*/
async function resolveMcpServerContext(req) {
try {
const appConfig = await getAppConfigForRequest(req);
return await resolveMCPServerContext({
mcpConfig: appConfig?.mcpConfig || {},
ensureConfigServers: (mcpConfig) => getMCPServersRegistry().ensureConfigServers(mcpConfig),
});
} catch (error) {
logger.warn(
'[resolveMcpServerContext] Failed to resolve MCP servers, degrading to empty:',
error,
);
return { configServers: {}, serverNames: [], rawServerNames: [] };
}
}
/**
* Resolves config-source servers and merges all server configs (YAML + config + user DB)
* for the given user context. Shared helper for controllers needing the full merged config.
* @param {string} userId
* @param {{ id?: string, role?: string }} [user]
* @returns {Promise<Record<string, import('@librechat/api').ParsedServerConfig>>}
*/
/**
* Names of every MCP server the user can reach (operator config + user DB),
* for legacy-key healing: collision detection in `initializeAgent` (consulted
* when a configured server name needs normalization) and the assistants heal
* in `healMcpToolNames` (always, since assistants reference user-owned
* servers too).
* @param {string} [userId]
* @param {string} [role]
* @returns {Promise<string[]>}
*/
async function getAccessibleMcpServerNames(userId, role) {
const configs = await resolveAllMcpConfigs(
userId,
role != null ? { id: userId, role } : { id: userId },
);
return Object.keys(configs ?? {});
}
/**
* Heals legacy raw-keyed MCP tool names in an assistant payload to the
* current normalized cache keys. Cached tool definitions are keyed
* `${toolName}${mcp_delimiter}${normalizeServerName(server)}`, while an
* assistant saved before that convention resubmits the raw-suffixed string
* on every edit — the controllers' exact-only lookup would then silently
* drop the tool from the assistant. SHADOWED raw names (normalized slot
* claimed by another configured server) stay raw and fail closed, mirroring
* the runtime heal, with the shadow set built from the FULL accessible
* audit (cross-tier collisions included) and healing skipped outright when
* that audit cannot complete. Config names are read only when a
* delimiter-bearing name actually misses the cache; config-read failures
* propagate (write path) rather than silently dropping the tool. Healed
* string entries dedupe order-preserving so a payload carrying both
* spellings can't submit duplicate function names.
* @param {object} params
* @param {ServerRequest} params.req
* @param {Array<string | object>} [params.tools]
* @param {Record<string, unknown>} params.toolDefinitions
* @returns {Promise<Array<string | object>>}
*/
async function healMcpToolNames({ req, tools, toolDefinitions, accessibleServerNames }) {
const list = tools ?? [];
const needsHeal = list.some(
(tool) =>
typeof tool === 'string' &&
tool.includes(Constants.mcp_delimiter) &&
toolDefinitions[tool] == null,
);
if (!needsHeal) {
return list;
}
/** Cross-tier shadowing (DB `foo` vs operator `foo!`) is invisible to
* operator names alone — the shadow set must come from the FULL
* accessible audit: assistants reference user-owned servers too (the
* definitions loader resolves them), so their pre-strip keys must heal
* against the same catalog. Callers holding the loader's snapshot pass
* it to avoid repeating the app-config and registry reads on the write
* path; without one, the audit is fetched here, and when it cannot
* complete healing is skipped entirely (the raw key stays raw and fails
* closed). */
let auditNames = accessibleServerNames;
if (auditNames == null) {
const rawServerNames = await resolveMcpConfigNames(req);
try {
const accessible = await getAccessibleMcpServerNames(req.user?.id, req.user?.role);
auditNames = [...new Set([...accessible, ...rawServerNames])];
} catch (error) {
logger.warn(
'[healMcpToolNames] Accessible-server audit unavailable; skipping legacy-key healing:',
error,
);
return list;
}
}
const shadowed = findShadowedServerNames(auditNames);
/** A pre-strip key persisted AFTER server-name normalization carries the
* NORMALIZED suffix, which the raw config names cannot match — the
* boundary must resolve against both spellings and map back to the raw
* name for the shadow and membership guards. */
const serverNameAliases = buildServerNameAliases(auditNames);
const boundaryNames = [...new Set([...auditNames, ...serverNameAliases.keys()])];
const seen = new Set();
const healedList = [];
for (const tool of list) {
let healedTool = tool;
if (
typeof tool === 'string' &&
tool.includes(Constants.mcp_delimiter) &&
toolDefinitions[tool] == null
) {
const [, parsedServerName] = splitMCPToolKey(tool, boundaryNames);
let rawServerName;
if (parsedServerName != null && auditNames.includes(parsedServerName)) {
rawServerName = parsedServerName;
} else if (parsedServerName != null) {
const aliased = serverNameAliases.get(parsedServerName);
/** A normalized spelling on a CONTESTED slot is ambiguous between the
* tie-break winner and its shadowed rivals — rewriting persisted
* data must fail closed here, mirroring the raw-spelling shadow
* guard, rather than bind the reference to the winner. */
const contested =
aliased != null &&
auditNames.some(
(name) => name !== aliased && normalizeServerName(name) === parsedServerName,
);
rawServerName = contested ? undefined : aliased;
}
if (rawServerName != null && !shadowed.has(rawServerName)) {
const healed = normalizeMCPToolKey(tool, auditNames);
if (toolDefinitions[healed] != null) {
healedTool = healed;
} else {
/** Catalog keys built after redundant-prefix stripping no longer
* match a pre-strip persisted key — without this second candidate
* the exact-lookup below silently drops the tool from the
* assistant. The rewrite only lands when the stripped key actually
* exists in the loaded definitions, so an unstripped catalog
* (collision guard kept the raw name) never heals into a phantom. */
const keyServerName = normalizeServerName(rawServerName);
const [healedToolName] = splitMCPToolKey(healed, [keyServerName]);
const strippedName = stripServerNamePrefix(healedToolName, keyServerName);
const strippedKey = `${strippedName}${Constants.mcp_delimiter}${keyServerName}`;
/** Rewrite only when the stripped entry PROVES the same upstream
* identity — a stale key for a removed tool must not be healed
* onto a different sibling that kept its raw name. */
if (
strippedName !== healedToolName &&
toolDefinitions[strippedKey]?.serverToolName === healedToolName
) {
healedTool = strippedKey;
}
}
}
}
/** A payload carrying both spellings collapses to one entry after the
* heal — duplicate function names make providers reject the save. */
if (typeof healedTool === 'string') {
if (seen.has(healedTool)) {
continue;
}
seen.add(healedTool);
}
healedList.push(healedTool);
}
return healedList;
}
/**
* Loads static and MCP function definitions used by assistant create/update writes. MCP catalogs
* are stored per server and effective config, so assistant writers must resolve the referenced
* server slices instead of relying on the static aggregate cache.
* @param {object} params
* @param {ServerRequest} params.req
* @param {ServerResponse} [params.res]
* @param {Array<string | object>} [params.tools]
* @returns {Promise<object>}
*/
async function getAssistantToolDefinitions({ req, res, tools }) {
const registry = getMCPServersRegistry();
const appConfig = await getAppConfigForRequest(req);
const oboIdentityContext = createAuthIdentityContext({
user: req.user,
tenantId: getTenantId(),
});
const upstreamTokenProvider = createOpenIDSessionTokenProvider({
req,
res: res ?? req.res,
user: req.user,
identityContext: oboIdentityContext,
tokenPreference: 'access_token',
});
return await loadAssistantToolDefinitions(
{
user: req.user,
tools,
staticTools: (await getCachedTools()) ?? {},
mcpConfig: appConfig?.mcpConfig ?? {},
},
{
ensureConfigServers: (mcpConfig) => registry.ensureConfigServers(mcpConfig),
getAllServerConfigs: (userId, configServers, role) =>
registry.getAllServerConfigs(userId, configServers, role),
getMCPServerTools,
getServerToolFunctionsSnapshot: async (userId, serverName, serverConfig) =>
(await getMCPManager()?.getServerToolFunctionsSnapshot(
userId,
serverName,
serverConfig,
)) ?? {
tools: null,
},
recoverServerTools: async (serverName, serverConfig) => {
const userMCPAuthMap = await getUserMCPAuthMap({
userId: req.user.id,
servers: [serverName],
findPluginAuthsByKeys,
});
const result = await reinitMCPServer({
user: req.user,
serverName,
serverConfig,
userMCPAuthMap,
upstreamTokenProvider,
oboIdentityContext,
});
return result?.availableTools ?? null;
},
cacheMCPServerTools,
},
);
}
/**
* Resolves the name set MCP collision guards audit against. Prefers the
* caller-threaded accessible set; self-fetches only when a configured name
* needs normalization (safe-name deployments never pay the lookup); reports
* `complete: false` when the full set was needed but unavailable — callers
* must then fail closed for normalization-sensitive references instead of
* auditing against operator names alone.
* @param {object} params
* @param {readonly string[]} params.rawServerNames
* @param {readonly string[]} [params.accessibleServerNames]
* @param {string} [params.userId]
* @param {string} [params.role]
* @returns {Promise<{ names: readonly string[], complete: boolean }>}
*/
async function resolveCollisionAuditNames({ rawServerNames, accessibleServerNames, userId, role }) {
if (accessibleServerNames?.length) {
return { names: accessibleServerNames, complete: true };
}
const needsFullAudit = rawServerNames.some((name) => normalizeServerName(name) !== name);
if (!needsFullAudit) {
return { names: rawServerNames, complete: true };
}
try {
const names = await getAccessibleMcpServerNames(userId, role);
/** `resolveAllMcpConfigs` tolerates `ensureConfigServers` failures, so
* the merged read can silently omit config-only servers. The caller's
* raw config names come from the app-config snapshot (registry-
* independent), so the union keeps `complete: true` honest. */
return { names: [...new Set([...names, ...rawServerNames])], complete: true };
} catch (error) {
logger.warn(
'[MCP] Collision audit unavailable; normalization-sensitive references fail closed:',
error,
);
return { names: rawServerNames, complete: false };
}
}
async function resolveAllMcpConfigs(userId, user) {
const registry = getMCPServersRegistry();
const appConfig = await getAppConfigForUser(userId, user);
let configServers = {};
try {
configServers = await registry.ensureConfigServers(appConfig?.mcpConfig || {});
} catch {
logger.warn('[resolveAllMcpConfigs] Config server resolution failed; continuing without');
}
if (user?.role) {
return await registry.getAllServerConfigs(userId, configServers, user.role);
}
return await registry.getAllServerConfigs(userId, configServers);
}
/**
* Best-effort early gate; the authoritative check is
* `assertResolvedRuntimeConfigAllowed` in `@librechat/api`, whose resolution
* this must mirror. Graph placeholders resolve later (async), so a URL still
* carrying one defers to the authoritative check instead of rejecting here.
*/
async function isEarlyDomainAllowed({
serverConfig,
user,
requestBody,
userMCPAuthMap,
serverName,
allowedDomains,
allowedAddresses,
}) {
const validationConfig = processMCPEnv({
user,
body: requestBody,
dbSourced: isUserSourced(serverConfig),
options: serverConfig,
customUserVars: getServerCustomUserVars(userMCPAuthMap, serverName),
});
if (
typeof validationConfig?.url === 'string' &&
containsGraphTokenPlaceholder(validationConfig.url)
) {
return true;
}
return await isMCPDomainAllowed(validationConfig, allowedDomains, allowedAddresses);
}
/**
* @param {string} toolName
* @param {string} serverName
*/
function createUnavailableToolStub(toolName, serverName) {
const normalizedToolKey = `${toolName}${Constants.mcp_delimiter}${normalizeServerName(serverName)}`;
const _call = async () => [unavailableMsg, null];
const toolInstance = tool(_call, {
schema: {
type: 'object',
properties: {
input: { type: 'string', description: 'Input for the tool' },
},
required: [],
},
name: normalizedToolKey,
description: unavailableMsg,
responseFormat: AgentConstants.CONTENT_AND_ARTIFACT,
});
toolInstance.mcp = true;
toolInstance.mcpRawServerName = serverName;
return toolInstance;
}
function isEmptyObjectSchema(jsonSchema) {
return (
jsonSchema != null &&
typeof jsonSchema === 'object' &&
jsonSchema.type === 'object' &&
(jsonSchema.properties == null || Object.keys(jsonSchema.properties).length === 0) &&
!jsonSchema.additionalProperties
);
}
/**
* @param {object} params
* @param {ServerResponse} params.res - The Express response object for sending events.
* @param {string} params.stepId - The ID of the step in the flow.
* @param {ToolCallChunk} params.toolCall - The tool call object containing tool information.
* @param {string | null} [params.streamId] - The stream ID for resumable mode.
* @param {number} [params.jobCreatedAt] - The generation epoch that owns emitted events.
*/
function createRunStepDeltaEmitter({ res, stepId, toolCall, streamId = null, jobCreatedAt }) {
/**
* @param {string} authURL - The URL to redirect the user for OAuth authentication.
* @param {{ expiresAt?: number }} [options]
* @returns {Promise<void>}
*/
return async function (authURL, options) {
const eventData = buildMCPAuthRunStepDeltaEvent({ authURL, stepId, toolCall, options });
if (streamId) {
await GenerationJobManager.emitChunk(streamId, eventData, {
expectedCreatedAt: jobCreatedAt,
});
} else {
sendEvent(res, eventData);
}
};
}
/**
* @param {object} params
* @param {ServerResponse} params.res - The Express response object for sending events.
* @param {string} params.runId - The Run ID, i.e. message ID
* @param {string} params.stepId - The ID of the step in the flow.
* @param {ToolCallChunk} params.toolCall - The tool call object containing tool information.
* @param {number} [params.index]
* @param {string | null} [params.streamId] - The stream ID for resumable mode.
* @param {number} [params.jobCreatedAt] - The generation epoch that owns emitted events.
* @returns {() => Promise<void>}
*/
function createRunStepEmitter({
res,
runId,
stepId,
toolCall,
index,
streamId = null,
jobCreatedAt,
}) {
return async function () {
const eventData = buildMCPAuthRunStepEvent({ runId, stepId, toolCall, index });
if (streamId) {
await GenerationJobManager.emitChunk(streamId, eventData, {
expectedCreatedAt: jobCreatedAt,
});
} else {
sendEvent(res, eventData);
}
};
}
/**
* Creates a function used to ensure the flow handler is only invoked once
* @param {object} params
* @param {string} params.flowId - The ID of the login flow.
* @param {FlowStateManager<any>} params.flowManager - The flow manager instance.
* @param {(authURL: string, options?: { expiresAt?: number }) => void | Promise<void>} [params.callback]
*/
function createOAuthStart({ flowId, flowManager, callback }) {
/**
* Creates a function to handle OAuth login requests.
* @param {string} authURL - The URL to redirect the user for OAuth authentication.
* @param {{ expiresAt?: number }} [options]
* @returns {Promise<boolean>} Returns true to indicate the event was sent successfully.
*/
return async function (authURL, options) {
let emitted = false;
const emitOAuthStart = async (message) => {
if (options) {
await callback?.(authURL, options);
} else {
await callback?.(authURL);
}
emitted = true;
logger.debug(message);
};
const existingFlow = await flowManager.getFlowState(flowId, 'oauth_login');
if (existingFlow) {
await emitOAuthStart('Re-sent OAuth login request to client');
return true;
}
await flowManager.createFlowWithHandler(flowId, 'oauth_login', async () => {
await emitOAuthStart('Sent OAuth login request to client');
return true;
});
if (!emitted) {
await emitOAuthStart('Re-sent OAuth login request to client');
}
return true;
};
}
/**
* @param {object} params
* @param {ServerResponse} params.res - The Express response object for sending events.
* @param {string} params.stepId - The ID of the step in the flow.
* @param {ToolCallChunk} params.toolCall - The tool call object containing tool information.
* @param {string | null} [params.streamId] - The stream ID for resumable mode.
* @param {number} [params.jobCreatedAt] - The generation epoch that owns emitted events.
*/
function createOAuthEnd({ res, stepId, toolCall, streamId = null, jobCreatedAt }) {
return async function () {
const eventData = buildMCPAuthRunStepEndDeltaEvent({ stepId, toolCall });
if (streamId) {
await GenerationJobManager.emitChunk(streamId, eventData, {
expectedCreatedAt: jobCreatedAt,
});
} else {
sendEvent(res, eventData);
}
logger.debug('Sent OAuth login success to client');
};
}
/**
* @param {Object} params
* @param {() => Promise<void>} params.runStepEmitter
* @param {(authURL: string, options?: { expiresAt?: number }) => Promise<void>} params.runStepDeltaEmitter
* @returns {(authURL: string, options?: { expiresAt?: number }) => Promise<void>}
*/
function createOAuthCallback({ runStepEmitter, runStepDeltaEmitter }) {
return async function (authURL, options) {
await runStepEmitter();
await runStepDeltaEmitter(authURL, options);
};
}
function resolveToolCallUserId({ effectiveUser, capturedUser, invocationUserId, serverConfig }) {
if (serverConfig?.obo == null) {
return effectiveUser?.id || invocationUserId || capturedUser?.id;
}
const effectiveUserId = effectiveUser?.id;
const capturedUserId = capturedUser?.id;
if (!effectiveUserId || !capturedUserId) {
throw new Error('OBO tool calls require matching captured and effective user ids');
}
if (effectiveUserId !== capturedUserId) {
throw new Error('OBO tool call user mismatch');
}
return effectiveUserId;
}
/**
* @param {Object} params
* @param {ServerResponse} params.res - The Express response object for sending events.
* @param {import('@librechat/api').UpstreamTokenProvider} [params.upstreamTokenProvider] - Live upstream-token closure for OBO, built at the request boundary so this layer never receives the raw Express request.
* @param {import('@librechat/api').AuthIdentityContext} [params.oboIdentityContext] - Non-template-visible OBO identity context built from the real request user.
* @param {IUser} params.user - The user from the request object.
* @param {string} params.serverName
* @param {AbortSignal} params.signal
* @param {string} params.model
* @param {number} [params.index]
* @param {string | null} [params.streamId] - The stream ID for resumable mode.
* @param {number} [params.jobCreatedAt] - The generation epoch that owns emitted events.
* @param {Record<string, Record<string, string>>} [params.userMCPAuthMap]
* @param {import('@librechat/api').RequestScopedMCPConnectionStore} [params.requestScopedConnections]
* @param {import('@librechat/api').ParsedServerConfig} [params.serverConfig] - Used to bypass reconnect throttling for request-scoped servers.
* @returns { Promise<Array<typeof tool | { _call: (toolInput: Object | string) => unknown}>> } An object with `_call` method to execute the tool input.
*/
async function reconnectServer({
res,
user,
index,
signal,
serverName,
serverConfig,
configServers,
userMCPAuthMap,
requestBody,
requestScopedConnections,
upstreamTokenProvider,
oboIdentityContext,
streamId = null,
jobCreatedAt,
}) {
logger.debug('[MCP][reconnectServer] Starting reconnect', {
userId: user?.id,
hasUserMCPAuthMap: Boolean(userMCPAuthMap),
});
// Request-scoped servers reconnect on every message by design; throttling them
// would stub out healthy tools for messages sent within the throttle window.
const requestScoped = serverConfig ? requiresEphemeralUserConnection(serverConfig) : false;
if (!requestScoped) {
const throttleKey = `${user.id}:${serverName}`;
const now = Date.now();
const lastAttempt = lastReconnectAttempts.get(throttleKey) ?? 0;
if (now - lastAttempt < RECONNECT_THROTTLE_MS) {
logger.debug('[MCP][reconnectServer] Throttled reconnect');
return null;
}
lastReconnectAttempts.set(throttleKey, now);
evictStale(lastReconnectAttempts, RECONNECT_THROTTLE_MS);
}
const runId = Constants.USE_PRELIM_RESPONSE_MESSAGE_ID;
const flowId = `${user.id}:${serverName}:${Date.now()}`;
const flowManager = getFlowStateManager(getLogStores(CacheKeys.FLOWS));
const stepId = buildMCPAuthStepId(serverName);
const toolCall = buildMCPAuthToolCall({
id: flowId,
serverName,
});
const runStepEmitter = createRunStepEmitter({
res,
index,
runId,
stepId,
toolCall,
streamId,
jobCreatedAt,
});
const runStepDeltaEmitter = createRunStepDeltaEmitter({
res,
stepId,
toolCall,
streamId,
jobCreatedAt,
});
const callback = createOAuthCallback({ runStepEmitter, runStepDeltaEmitter });
const oauthStart = createOAuthStart({
res,
flowId,
callback,
flowManager,
});
return await reinitMCPServer({
user,
signal,
serverName,
configServers,
oauthStart,
flowManager,
userMCPAuthMap,
requestBody,
requestScopedConnections,
upstreamTokenProvider,
oboIdentityContext,
forceNew: true,
returnOnOAuth: false,
connectionTimeout: Time.THIRTY_SECONDS,
});
}
/**
* Creates all tools from the specified MCP Server via `toolKey`.
*
* This function assumes tools could not be aggregated from the cache of tool definitions,
* i.e. `availableTools`, and will reinitialize the MCP server to ensure all tools are generated.
*
* @param {Object} params
* @param {ServerResponse} params.res - The Express response object for sending events.
* @param {{ canUseServers: (user?: IUser) => Promise<boolean> }} [params.mcpPermissionContext] - Request-scoped MCP permission context.
* @param {IUser} params.user - The user from the request object.
* @param {string} params.serverName
* @param {string} params.model
* @param {Providers | EModelEndpoint} params.provider - The provider for the tool.
* @param {number} [params.index]
* @param {AbortSignal} [params.signal]
* @param {string | null} [params.streamId] - The stream ID for resumable mode.
* @param {number} [params.jobCreatedAt] - The generation epoch that owns emitted events.
* @param {import('@librechat/api').ParsedServerConfig} [params.config]
* @param {import('@librechat/api').RequestBody} [params.requestBody]
* @param {import('@librechat/api').RequestScopedMCPConnectionStore} [params.requestScopedConnections]
* @param {Record<string, Record<string, string>>} [params.userMCPAuthMap]
* @param {import('@librechat/api').UpstreamTokenProvider} [params.upstreamTokenProvider] - Live upstream-token closure for OBO, built at the request boundary.
* @param {import('@librechat/api').AuthIdentityContext} [params.oboIdentityContext] - Non-template-visible OBO identity context built from the real request user.
* @returns { Promise<Array<typeof tool | { _call: (toolInput: Object | string) => unknown}>> } An object with `_call` method to execute the tool input.
*/
async function createMCPTools({
res,
mcpPermissionContext,
user,
index,
signal,
config,
provider,
serverName,
configServers,
userMCPAuthMap,
requestBody,
requestScopedConnections,
upstreamTokenProvider,
oboIdentityContext,
streamId = null,
jobCreatedAt,
}) {
const serverConfig =
config ?? (await getMCPServersRegistry().getServerConfig(serverName, user?.id, configServers));
if (serverConfig?.url) {
const appConfig = await getAppConfig({
role: user?.role,
tenantId: user?.tenantId,
userId: user?.id,
});
const allowedDomains = appConfig?.mcpSettings?.allowedDomains;
const allowedAddresses = appConfig?.mcpSettings?.allowedAddresses;
const isDomainAllowed = await isEarlyDomainAllowed({
serverConfig,
user,
requestBody,
userMCPAuthMap,
serverName,
allowedDomains,
allowedAddresses,
});
if (!isDomainAllowed) {
logger.warn('[MCP] Domain not allowed; skipping all server tools');
return [];
}
}
const result = await reconnectServer({
res,
user,
index,
signal,
serverName,
serverConfig,
configServers,
userMCPAuthMap,
requestBody,
requestScopedConnections,
upstreamTokenProvider,
oboIdentityContext,
streamId,
jobCreatedAt,
});
if (result === null) {
logger.debug('[MCP] Reconnect throttled; skipping tool creation');
return [];
}
if (!result || !result.tools) {
logger.warn('[MCP] Failed to reinitialize server');
return [];
}
const serverTools = [];
const keyServerName = normalizeServerName(serverName);
const keyToolNames = stripServerNamePrefixes(
result.tools.map((tool) => tool.name),
keyServerName,
);
for (const tool of result.tools) {
const toolInstance = await createMCPTool({
res,
mcpPermissionContext,
user,
provider,
userMCPAuthMap,
configServers,
streamId,
jobCreatedAt,
availableTools: result.availableTools,
serverName,
/** Model-facing key: matches the normalized `availableTools` keys and
* the instance name `createToolInstance` will assign. */
toolKey: `${keyToolNames.get(tool.name) ?? tool.name}${Constants.mcp_delimiter}${keyServerName}`,
requestBody,
requestScopedConnections,
upstreamTokenProvider,
oboIdentityContext,
config: serverConfig,
});
if (toolInstance) {
serverTools.push(toolInstance);
}
}
return serverTools;
}
/**
* Creates a single tool from the specified MCP Server via `toolKey`.
* @param {Object} params
* @param {ServerResponse} params.res - The Express response object for sending events.
* @param {{ canUseServers: (user?: IUser) => Promise<boolean> }} [params.mcpPermissionContext] - Request-scoped MCP permission context.
* @param {IUser} params.user - The user from the request object.
* @param {string} params.toolKey - The toolKey for the tool.
* @param {string} params.model - The model for the tool.
* @param {number} [params.index]
* @param {AbortSignal} [params.signal]
* @param {string | null} [params.streamId] - The stream ID for resumable mode.
* @param {Providers | EModelEndpoint} params.provider - The provider for the tool.
* @param {LCAvailableTools} [params.availableTools]
* @param {import('@librechat/api').RequestBody} [params.requestBody]
* @param {import('@librechat/api').RequestScopedMCPConnectionStore} [params.requestScopedConnections]
* @param {Record<string, Record<string, string>>} [params.userMCPAuthMap]
* @param {import('@librechat/api').ParsedServerConfig} [params.config]
* @param {import('@librechat/api').UpstreamTokenProvider} [params.upstreamTokenProvider] - Live upstream-token closure for OBO, built at the request boundary.
* @param {import('@librechat/api').AuthIdentityContext} [params.oboIdentityContext] - Non-template-visible OBO identity context built from the real request user.
* @param {string} [params.serverName] - Resolved raw MCP server name from tool loading.
* @param {(availableTools: LCAvailableTools) => void} [params.onAvailableTools]
* @param {number} [params.jobCreatedAt] - The generation epoch that owns emitted events.
* @returns { Promise<typeof tool | { _call: (toolInput: Object | string) => unknown}> } An object with `_call` method to execute the tool input.
*/
async function createMCPTool({
res,
mcpPermissionContext,
user,
index,
signal,
toolKey,
provider,
userMCPAuthMap,
availableTools,
requestBody,
requestScopedConnections,
config,
configServers,
upstreamTokenProvider,
oboIdentityContext,
serverName: resolvedServerName,
onAvailableTools,
streamId = null,
jobCreatedAt,
}) {
/** `loadTools` already resolved the server for this key; parsing is the fallback. */
const [parsedToolName, parsedServerName] = splitMCPToolKey(
toolKey,
/** Current keys embed the NORMALIZED server name, legacy persisted keys
* the RAW one — the candidate list needs both spellings or a raw name
* that contains the delimiter mis-splits under the generic fallback. */
resolvedServerName
? [resolvedServerName, normalizeServerName(resolvedServerName)]
: Object.keys(configServers ?? {}).flatMap((name) => [name, normalizeServerName(name)]),
);
let serverName = resolvedServerName ?? parsedServerName;
const toolName = parsedToolName;
let serverConfig =
config ?? (await getMCPServersRegistry().getServerConfig(serverName, user?.id, configServers));
/** DIRECT-FIRST alias fallback: only when the parsed name resolves to no
* server is it treated as the normalized spelling of a raw config name —
* a user-DB server named like an operator server's normalized form must
* keep its own identity. */
if (!serverConfig && resolvedServerName == null && parsedServerName != null) {
const aliasedName = buildServerNameAliases(Object.keys(configServers ?? {})).get(
parsedServerName,
);
if (aliasedName != null && aliasedName !== parsedServerName) {
serverConfig = await getMCPServersRegistry().getServerConfig(
aliasedName,
user?.id,
configServers,
);
if (serverConfig) {
serverName = aliasedName;
}
}
}
const requestScopedTools = serverConfig ? requiresEphemeralUserConnection(serverConfig) : false;
const useMissingToolCache = !requestScopedTools;
if (serverConfig?.url) {
const appConfig = await getAppConfig({
role: user?.role,
tenantId: user?.tenantId,
userId: user?.id,
});
const allowedDomains = appConfig?.mcpSettings?.allowedDomains;
const allowedAddresses = appConfig?.mcpSettings?.allowedAddresses;
const isDomainAllowed = await isEarlyDomainAllowed({
serverConfig,
user,
requestBody,
userMCPAuthMap,
serverName,
allowedDomains,
allowedAddresses,
});
if (!isDomainAllowed) {
logger.warn('[MCP] Domain no longer allowed; skipping tool creation');
return undefined;
}
}
/** Legacy keys persisted pre-normalization (assistants, direct tool
* calls) carry the RAW server name, while `availableTools` is keyed by
* the canonical normalized key — look up both spellings. Keys are also
* built after redundant server-name-prefix stripping now, so a persisted
* pre-strip key (`acme_foo_mcp_acme`) must additionally try
* its stripped spelling or the tool degrades to an unavailable stub. */
const keyServerName = serverName != null ? normalizeServerName(serverName) : undefined;
const canonicalToolKey =
keyServerName != null ? `${toolName}${Constants.mcp_delimiter}${keyServerName}` : toolKey;
const strippedToolName =
keyServerName != null ? stripServerNamePrefix(toolName, keyServerName) : toolName;
const strippedToolKey =
strippedToolName !== toolName
? `${strippedToolName}${Constants.mcp_delimiter}${keyServerName}`
: null;
const candidateToolKeys = [toolKey];
if (canonicalToolKey !== toolKey) {
candidateToolKeys.push(canonicalToolKey);
}
if (strippedToolKey != null && !candidateToolKeys.includes(strippedToolKey)) {
candidateToolKeys.push(strippedToolKey);
}
let matchedToolKey = toolKey;
const findToolEntry = (tools) => {
for (const key of candidateToolKeys) {
const entry = tools?.[key];
if (!entry?.function) {
continue;
}
/** The stripped-spelling candidate is only a legacy match when the
* entry PROVES the same upstream identity — without this, a stale
* reference to a removed tool could strip onto a DIFFERENT sibling
* that kept its raw name and silently call the wrong tool. */
if (key === strippedToolKey && entry.serverToolName !== toolName) {
continue;
}
matchedToolKey = key;
return entry;
}
return undefined;
};
/** @type {LCFunctionTool | undefined} */
let toolEntry = findToolEntry(availableTools);
if (!toolEntry) {
const cachedAt = useMissingToolCache ? missingToolCache.get(toolKey) : undefined;
if (cachedAt && Date.now() - cachedAt < MISSING_TOOL_TTL_MS) {
logger.debug('[MCP] Tool is in negative cache; returning unavailable stub');
return createUnavailableToolStub(toolName, serverName);
}
logger.warn('[MCP] Requested tool not found in available tools; reinitializing server');
const result = await reconnectServer({
res,
user,
index,
signal,
serverName,
serverConfig,
configServers,
userMCPAuthMap,
requestBody,
requestScopedConnections,
upstreamTokenProvider,
oboIdentityContext,
streamId,
jobCreatedAt,
});
if (result?.availableTools) {
onAvailableTools?.(result.availableTools);
}
toolEntry = findToolEntry(result?.availableTools);
if (!toolEntry && useMissingToolCache) {
missingToolCache.set(toolKey, Date.now());
evictStale(missingToolCache, MISSING_TOOL_TTL_MS);
}
}
if (!toolEntry) {
logger.warn(
`[MCP][${serverName}][${toolName}] Tool definition not found, returning unavailable stub.`,
);
return createUnavailableToolStub(toolName, serverName);
}
return createToolInstance({
res,
mcpPermissionContext,
user,
requestBody,
requestScopedConnections,
provider,
/** A legacy pre-strip key that resolves to the stripped entry KEEPS its
* persisted spelling as the instance name: `agent.tools` entries and
* `tool_options` keys reference that spelling, and renaming the instance
* would silently detach those per-tool settings. The upstream call name
* still comes from the MATCHED entry — its recorded raw name, or the
* matched key's own tool half when the entry was never stripped. */
toolName,
serverToolName:
toolEntry.serverToolName ??
(matchedToolKey === strippedToolKey ? strippedToolName : toolName),
currentToolName: matchedToolKey === strippedToolKey ? strippedToolName : undefined,
serverName,
serverConfig,
toolDefinition: toolEntry['function'],
upstreamTokenProvider,
oboIdentityContext,
streamId,
jobCreatedAt,
});
}
function createToolInstance({
res,
mcpPermissionContext,
user: capturedUser = null,
requestBody: capturedRequestBody,
requestScopedConnections: capturedRequestScopedConnections,
toolName,
serverToolName = toolName,
currentToolName,
serverName,
serverConfig: capturedServerConfig,
toolDefinition,
provider: capturedProvider,
upstreamTokenProvider: capturedUpstreamTokenProvider = null,
oboIdentityContext: capturedOboIdentityContext = null,
streamId = null,
jobCreatedAt,
}) {
/** @type {LCTool} */
const { description, parameters } = toolDefinition;
const isGoogle = capturedProvider === Providers.VERTEXAI || capturedProvider === Providers.GOOGLE;
let schema = parameters ? normalizeJsonSchema(resolveJsonSchemaRefs(parameters)) : null;
if (schema && isGoogle) {
// Gemini/Vertex AI accept only a subset of JSON Schema; sanitize so MCP tools with
// unions, non-string enums, etc. don't 400 (they work as-is on OpenAI/Claude).
schema = sanitizeGeminiSchema(schema);
}
if (!schema || (isGoogle && isEmptyObjectSchema(schema))) {
schema = {
type: 'object',
properties: {
input: { type: 'string', description: 'Input for the tool' },
},
required: [],
};
}
const normalizedToolKey = `${toolName}${Constants.mcp_delimiter}${normalizeServerName(serverName)}`;
/** @type {(toolArguments: Object | string, config?: GraphRunnableConfig) => Promise<unknown>} */
const _call = async (toolArguments, config) => {
const effectiveUser = config?.configurable?.user ?? capturedUser;
const permissionUser = effectiveUser;
/** @type {string | undefined} */
let userId;
try {
userId = resolveToolCallUserId({
effectiveUser,
capturedUser,
invocationUserId: config?.configurable?.user_id,
serverConfig: capturedServerConfig,
});
const provider = (config?.metadata?.provider || capturedProvider)?.toLowerCase();
const canUseMCP = mcpPermissionContext
? await mcpPermissionContext.canUseServers(permissionUser)
: await userCanUseMCPServers(permissionUser);
if (!canUseMCP) {
throw new Error('Forbidden: Insufficient MCP server permissions');
}
const flowsCache = getLogStores(CacheKeys.FLOWS);
const flowManager = getFlowStateManager(flowsCache);
const derivedSignal = config?.signal ? AbortSignal.any([config.signal]) : undefined;
const mcpManager = getMCPManager(userId);
const { args: _args, stepId, ...toolCall } = config.toolCall ?? {};
const flowId = `${serverName}:oauth_login:${config.metadata.thread_id}:${config.metadata.run_id}`;
const runStepDeltaEmitter = createRunStepDeltaEmitter({
res,
stepId,
toolCall,
streamId,
jobCreatedAt,
});
const oauthStart = createOAuthStart({
flowId,
flowManager,
callback: runStepDeltaEmitter,
});
const oauthEnd = createOAuthEnd({
res,
stepId,
toolCall,
streamId,
jobCreatedAt,
});
const customUserVars =
config?.configurable?.userMCPAuthMap?.[`${Constants.mcp_prefix}${serverName}`];
/**
* The upstream-token closure is built at the request boundary (where
* `req`/`res` are in scope) and captured here, so this layer never holds
* the raw Express request. The closure reads/refreshes the LIVE
* `req.session.openidTokens` at call time and persists rotations; it is a
* no-op when reuse is off or the user is non-OpenID. A browser request whose session loses
* openidTokens rejects instead of falling back to a stale strategy snapshot.
* `tokenPreference: 'access_token'` (set at construction)
* is required for OBO since the grant sends the access token to the IdP
* as the jwt-bearer assertion.
*/
const result = await mcpManager.callTool({
serverName,
serverConfig: capturedServerConfig,
/** The upstream server never sees stripped names — a key that dropped
* a redundant server-name prefix calls the ORIGINAL tool. */
toolName: serverToolName,
provider,
toolArguments,
options: {
signal: derivedSignal,
},
user: effectiveUser,
requestBody: config?.configurable?.requestBody ?? capturedRequestBody,
requestScopedConnections:
config?.configurable?.requestScopedConnections ?? capturedRequestScopedConnections,
customUserVars,
flowManager,
tokenMethods: {
findToken,
createToken,
updateToken,
deleteTokens,
},
oauthStart,
oauthEnd,
graphTokenResolver: getGraphApiToken,
oboTokenResolver: exchangeOboToken,
oboTrustChecker: createOboTrustChecker(),
upstreamTokenProvider: capturedUpstreamTokenProvider,
oboIdentityContext: capturedOboIdentityContext,
});
if (isAssistantsEndpoint(provider) && Array.isArray(result)) {
return result[0];
}
return result;
} catch (error) {
logger.error(
`[MCP][${serverName}][${toolName}][User: ${userId}] Error calling MCP tool:`,
error,
);
/** Carries the actionable re-auth message; the substring heuristic below would misreport it as an OAuth configuration problem */
if (error instanceof OpenIDReauthRequiredError) {
throw error;
}
/** OAuth error, provide a helpful message */
const isOAuthError =
error.message?.includes('401') ||
error.message?.includes('OAuth') ||
error.message?.includes('authentication') ||
error.message?.includes('Non-200 status code (401)');
const isOAuthFlowSignal =
error.message === 'OAuth flow initiated - return early' ||
error.message === 'Pending OAuth flow reused - return early';
if (isOAuthError) {
if (
capturedServerConfig &&
!requiresOAuthMachinery(capturedServerConfig) &&
!isOAuthFlowSignal
) {
throw new Error(
`[MCP][${serverName}][${toolName}] upstream authentication failed; MCP OAuth is not configured for this server.`,
);
}
throw new Error(
`[MCP][${serverName}][${toolName}] OAuth authentication required. Please check the server logs for the authentication URL.`,
);
}
throw new Error(
`[MCP][${serverName}][${toolName}] tool call failed${error?.message ? `: ${error?.message}` : '.'}`,
);
}
};
const toolInstance = tool(_call, {
schema,
name: normalizedToolKey,
description: description || '',
responseFormat: AgentConstants.CONTENT_AND_ARTIFACT,
});
toolInstance.mcp = true;
toolInstance.mcpRawServerName = serverName;
if (serverToolName !== toolName) {
/** Upstream identity for stripped keys — lets the options aliasing in
* `buildToolClassification` heal legacy `tool_options` spellings. */
toolInstance.mcpServerToolName = serverToolName;
}
if (currentToolName != null && currentToolName !== toolName) {
/** Current catalog spelling for a LEGACY-named instance, so approval
* policies and hook matchers written against the current name still
* reach it (see `collectMCPToolAliases`). */
toolInstance.mcpCurrentToolName = currentToolName;
}
// Ephemeral request-scoped servers (runtime body placeholders) tear their
// connection down at request end, so they must never be backgrounded. A
// missing/stale config means the server's lifetime is unknowable, so fail
// closed (foreground) rather than risk a detached call against a torn-down
// connection.
toolInstance.mcpRequiresEphemeralConnection = capturedServerConfig
? requiresEphemeralUserConnection(capturedServerConfig)
: true;
// On Google/Vertex, propagate the union-flattened schema so definitions extracted
// from this instance don't reach the Gemini converter with unsupported unions.
toolInstance.mcpJsonSchema = isGoogle ? schema : parameters;
return toolInstance;
}
/**
* Get MCP setup data including config, connections, and OAuth servers.
* Resolves config-source servers from admin Config overrides when tenant context is available.
* @param {string} userId - The user ID
* @param {{ role?: string, tenantId?: string }} [options] - Optional role/tenant context
* @returns {Object} Object containing mcpConfig, appConnections, userConnections, and oauthServers
*/
async function getMCPSetupData(userId, options = {}) {
const registry = getMCPServersRegistry();
const { role, tenantId } = options;
const appConfig = await getAppConfig({ role, tenantId, userId });
const configServers = await registry.ensureConfigServers(appConfig?.mcpConfig || {});
const mcpConfig = role
? await registry.getAllServerConfigs(userId, configServers, role)
: await registry.getAllServerConfigs(userId, configServers);
const mcpManager = getMCPManager(userId);
/** @type {Map<string, import('@librechat/api').MCPConnection>} */
let appConnections = new Map();
try {
// Use getLoaded() instead of getAll() to avoid forcing connection creation.
// getAll() creates connections for all servers, which is problematic for servers
// that require user context (e.g., those with {{LIBRECHAT_USER_ID}} placeholders).
appConnections = (await mcpManager.appConnections?.getLoaded()) || new Map();
} catch (error) {
logger.error(`[MCP][User: ${userId}] Error getting app connections:`, error);
}
const userConnections = mcpManager.getUserConnections(userId) || new Map();
const oauthServers = new Set(
Object.entries(mcpConfig)
.filter(([, config]) => isOAuthServer(config))
.map(([name]) => name),
);
return {
mcpConfig,
oauthServers,
appConnections,
userConnections,
};
}
/**
* Check OAuth flow status for a user and server
* @param {string} userId - The user ID
* @param {string} serverName - The server name
* @param {string} [tenantId] - The tenant ID for the current request.
* @returns {Object} Object containing active and failed flow flags
*/
async function checkOAuthFlowStatus(userId, serverName, tenantId = getTenantId()) {
const flowsCache = getLogStores(CacheKeys.FLOWS);
const flowManager = getFlowStateManager(flowsCache);
const flowId = getOAuthFlowId(userId, serverName, tenantId);
try {
const flowState = await flowManager.getFlowState(flowId, 'mcp_oauth');
if (!flowState) {
return { hasActiveFlow: false, hasFailedFlow: false };
}
const flowAge = Date.now() - flowState.createdAt;
// Report active only while the flow is still usable (the handling/reuse window),
// not for the full Keyv retention TTL — otherwise the UI shows "connecting" for a
// flow the initiate/callback paths already reject, hiding the connect button.
const flowTTL = flowState.ttl || PENDING_STALE_MS;
if (flowState.status === 'FAILED' || (flowState.status === 'PENDING' && flowAge > flowTTL)) {
const wasCancelled = /abort|cancel/i.test(flowState.error ?? '');
if (wasCancelled) {
logger.debug(`[MCP Connection Status] Found cancelled OAuth flow for ${serverName}`, {
flowId,
status: flowState.status,
error: flowState.error,
});
return { hasActiveFlow: false, hasFailedFlow: false };
} else {
logger.debug(`[MCP Connection Status] Found failed OAuth flow for ${serverName}`, {
flowId,
status: flowState.status,
flowAge,
flowTTL,
timedOut: flowAge > flowTTL,
error: flowState.error,
});
return { hasActiveFlow: false, hasFailedFlow: true };
}
}
if (flowState.status === 'PENDING') {
logger.debug(`[MCP Connection Status] Found active OAuth flow for ${serverName}`, {
flowId,
flowAge,
flowTTL,
});
return { hasActiveFlow: true, hasFailedFlow: false };
}
return { hasActiveFlow: false, hasFailedFlow: false };
} catch (error) {
logger.error(`[MCP Connection Status] Error checking OAuth flows for ${serverName}:`, error);
return { hasActiveFlow: false, hasFailedFlow: false };
}
}
async function hasDurableMCPAuthorization(userId, serverName, config, runtimeContext = {}) {
const userMCPAuthMap =
runtimeContext.userMCPAuthMap ?? (await runtimeContext.loadUserMCPAuthMap?.());
const customUserVars = getServerCustomUserVars(userMCPAuthMap, serverName);
if (getMissingCustomUserVars(config, customUserVars).length > 0) {
return false;
}
const dbSourced = isUserSourced(config);
const bindingConfig = {
...config,
args: undefined,
env: undefined,
headers: undefined,
oauth_headers: undefined,
};
const graphProcessedConfig = dbSourced
? bindingConfig
: await preProcessGraphTokens(bindingConfig, {
user: runtimeContext.user,
graphTokenResolver: getGraphApiToken,
scopes: process.env.GRAPH_API_SCOPES,
});
const runtimeConfig = processMCPEnv({
user: runtimeContext.user,
options: graphProcessedConfig,
dbSourced,
customUserVars,
});
const allowlists = await (runtimeContext.loadMCPAllowlists?.() ??
getMCPServersRegistry().resolveAllowlists({
userId,
role: runtimeContext.user?.role,
}));
if (
runtimeConfig.url &&
!(await isMCPDomainAllowed(
runtimeConfig,
allowlists.allowedDomains,
allowlists.allowedAddresses,
))
) {
return false;
}
return MCPTokenStorage.hasStoredAuthorization({
userId,
serverName,
findToken,
validateClientBinding: (clientInfo, storedMetadata) =>
MCPOAuthHandler.assertStoredClientBinding(
serverName,
runtimeConfig.url,
clientInfo,
storedMetadata,
runtimeConfig.oauth,
),
});
}
async function getMCPUserConfigurationState(serverName, config, runtimeContext = {}) {
if (!hasCustomUserVars(config)) {
return undefined;
}
const userMCPAuthMap =
runtimeContext.userMCPAuthMap ?? (await runtimeContext.loadUserMCPAuthMap?.());
const customUserVars = getServerCustomUserVars(userMCPAuthMap, serverName);
return getMissingCustomUserVars(config, customUserVars).length > 0
? 'needs_configuration'
: 'configured';
}
function canDetectMCPRuntimeOAuth(config) {
return config.requiresOAuth == null && config.apiKey == null && hasRuntimeUrlPlaceholders(config);
}
/**
* Get connection status for a specific MCP server
* @param {string} userId - The user ID
* @param {string} serverName - The server name
* @param {import('@librechat/api').ParsedServerConfig} config - The server configuration
* @param {Map<string, import('@librechat/api').MCPConnection>} appConnections - App-level connections
* @param {Map<string, import('@librechat/api').MCPConnection>} userConnections - User-level connections
* @param {Set} oauthServers - Set of OAuth servers
* @param {{ user?: Partial<IUser>, userMCPAuthMap?: Record<string, Record<string, string>>, loadUserMCPAuthMap?: () => Promise<Record<string, Record<string, string>> | undefined>, loadMCPAllowlists?: () => Promise<{ allowedDomains?: string[] | null, allowedAddresses?: string[] | null }> }} [runtimeContext]
* @returns {Object} Object containing requiresOAuth, requestScoped, connectionState, and authorizationState
*/
async function getServerConnectionStatus(
userId,
serverName,
config,
appConnections,
userConnections,
oauthServers,
runtimeContext = {},
) {
const connection = appConnections.get(serverName) || userConnections.get(serverName);
const isStaleOrDoNotExist = connection ? connection?.isStale(config.updatedAt) : true;
const configuredOAuth = oauthServers.has(serverName);
const liveConnectionOAuth = connection?.usesOAuth?.() === true;
const runtimeOAuthCandidate = canDetectMCPRuntimeOAuth(config);
const effectiveOAuth = configuredOAuth || liveConnectionOAuth;
const requestScoped = requiresEphemeralUserConnection(config);
const configurationState = requestScoped
? await getMCPUserConfigurationState(serverName, config, runtimeContext)
: undefined;
const baseConnectionState = isStaleOrDoNotExist
? 'disconnected'
: connection?.connectionState || 'disconnected';
let finalConnectionState = baseConnectionState;
let requiresOAuth = effectiveOAuth;
let authorizationState = effectiveOAuth ? 'needs_authorization' : 'not_required';
// connection state overrides specific to OAuth servers
if (effectiveOAuth && baseConnectionState === 'connected') {
authorizationState = 'authorized';
} else if (effectiveOAuth && baseConnectionState === 'connecting') {
authorizationState = 'authorizing';
} else if (effectiveOAuth && baseConnectionState === 'error') {
authorizationState = 'error';
} else if (baseConnectionState === 'disconnected' && (effectiveOAuth || runtimeOAuthCandidate)) {
// check if server is actively being reconnected
const oauthReconnectionManager = getOAuthReconnectionManager();
if (oauthReconnectionManager.isReconnecting(userId, serverName)) {
requiresOAuth = true;
finalConnectionState = 'connecting';
authorizationState = 'authorizing';
} else {
const { hasActiveFlow, hasFailedFlow } = await checkOAuthFlowStatus(userId, serverName);
if (hasFailedFlow) {
requiresOAuth = true;
finalConnectionState = 'error';
authorizationState = 'error';
} else if (hasActiveFlow) {
requiresOAuth = true;
finalConnectionState = 'connecting';
authorizationState = 'authorizing';
} else if (await hasDurableMCPAuthorization(userId, serverName, config, runtimeContext)) {
/** OAuth readiness is durable even when this pod has no live connection. */
requiresOAuth = true;
finalConnectionState = 'connected';
authorizationState = 'authorized';
}
}
}
return {
requiresOAuth,
...(requestScoped && { requestScoped: true }),
...(configurationState && { configurationState }),
connectionState: finalConnectionState,
authorizationState,
};
}
module.exports = {
createMCPTool,
createMCPTools,
toProviderToolDefinition,
createMCPPermissionContext,
userCanUseMCPServers,
getMCPSetupData,
resolveConfigServers,
resolveMcpServerNames,
resolveMcpServerContext,
getAccessibleMcpServerNames,
healMcpToolNames,
getAssistantToolDefinitions,
resolveCollisionAuditNames,
resolveMcpConfigNames,
resolveAllMcpConfigs,
createOAuthStart,
checkOAuthFlowStatus,
getServerConnectionStatus,
createUnavailableToolStub,
};