different-ai/openwork · critical · EnterpriseMcpClientError

MCP_AUTHORIZATION_CALLBACK_FAILED

MCP_AUTHORIZATION_CALLBACK_FAILED

Error message

Enterprise MCP failed during ${phaseLabel[input.operationPhase]}${request}.

What it means

In completeAuthorization(), the authorization code was exchanged (tokens saved) but the subsequent validation (protocol handshake / tools/list) failed, AND the cleanup attempt to invalidate the just-exchanged credentials via credentialPort.invalidate() also failed. The client wraps both failures in an AggregateError inside an EnterpriseMcpClientError with operationPhase "authorization-callback" and code MCP_AUTHORIZATION_CALLBACK_FAILED. This is serious: possibly-orphaned OAuth credentials remain persisted and unvalidated.

Source

Thrown at packages/enterprise-mcp-client/src/enterprise-mcp-client.ts:572

            operationFailed = true
            let credentialCleanupError: unknown = null
            if (exchangedTokens && credentialPort) {
              try {
                const cleanupController = new AbortController()
                await credentialPort.invalidate({
                  context: {
                    connectionId: input.connection.id,
                    commitExpiresAt: clock.now() + closeTimeoutMs,
                    signal: cleanupController.signal,
                  },
                  reason: "post-authorization-validation-failed",
                })
              } catch (cleanupError) {
                credentialCleanupError = cleanupError
              }
            }
            if (credentialCleanupError) {
              throw new EnterpriseMcpClientError({
                operationPhase: "authorization-callback",
                requestPhase: session.observer.lastRequestPhase(),
                cause: new AggregateError(
                  [error, credentialCleanupError],
                  "Post-authorization validation failed and the exchanged credentials could not be invalidated.",
                ),
              })
            }
            throw error
          } finally {
            try {
              await closeWithinDeadline(() => session.client.close(), closeTimeoutMs)
            } catch (error) {
              if (!operationFailed) {
                throw new EnterpriseMcpClientError({
                  operationPhase: "shutdown",
                  requestPhase: session.observer.lastRequestPhase(),
                  cause: error,

View on GitHub (pinned to 2b7df46e8a)

Solutions

Inspect the AggregateError's two causes: the primary validation failure and the cleanup failure; fix each accordingly.
Manually invalidate the exchanged credentials for this connectionId in your persistence store to avoid orphaned tokens.
Check the persistence adapter's health and its commit deadline handling; increase the lifecycle window if it expired mid-callback.
Retry completeAuthorization after cleanup is possible; the exchange may need to be redone from a fresh authorization code.

Defensive patterns

Strategy: try-catch

Validate before calling

async function assertPersistenceHealthy(credentialsPort) {
  // cheap probe that the credential store accepts writes before exchanging the code
  await credentialsPort.ping?.();
}

Type guard

function isAggregateCleanupError(e: unknown): e is AggregateError {
  return e instanceof AggregateError && e.message.includes("could not be invalidated");
}

Try / catch

try {
  await client.completeAuthorization(input);
} catch (e) {
  if (e.code === "MCP_AUTHORIZATION_CALLBACK_FAILED") {
    const [validationError, cleanupError] = e.cause?.errors ?? [];
    // 1) manually invalidate credentials for connectionId, 2) restart OAuth from a fresh code
  }
  throw e;
}

Prevention

Monitor the persistence backend's health before running OAuth callbacks.
Keep the lifecycle deadline long enough to cover exchange + validation + cleanup.
Reconcile orphaned credentials periodically (tokens persisted without a validated session).
Alert on MCP_AUTHORIZATION_CALLBACK_FAILED — it implies two simultaneous failures.

When it happens

Trigger: finishAuth(code) succeeds, then connectWithProtocolNegotiation or listTools throws (e.g. server 401/5xx/network drop), then persistence.credentials.invalidate({ reason: "post-authorization-validation-failed" }) also throws (persistence backend down, commit deadline expired, signal aborted).

Common situations: Persistence adapter (database) outage coinciding with a failing MCP server; lifecycle deadline expiring mid-callback; the OAuth adapter rejecting the invalidation because the transaction already committed or the signal was aborted.

Related errors

AI-assisted analysis of different-ai/openwork@2b7df46e8a (2026-09-01). Data as JSON: /api/errors/f2008e1a73c0b01a. Report an issue: GitHub.