Repository navigation
fix(mcp): stop blocking connections for sign-in server review - #7456
Closed
reachjalil wants to merge 1 commit into
Closed
reachjalil wants to merge 1 commit into
reachjalil wants to merge 1 commit into
Conversation
An OAuth connection could be marked as needing an administrator's review of its sign-in server. While marked, search reported it as an error, tool calls and the per-connection routes refused it, and nobody could use it until an admin chose an authorization server in its settings. The mark was meant for a provider whose sign-in server changed. It was set by any sign-in configuration error while saving discovery, on the first reading, on connections nobody had signed in to, and it never cleared by itself. It held connections that had never worked and left them unusable after the provider's sign-in information was consistent again. It was also not what kept credentials safe. The client library checks, every time it loads or saves discovery, that the authorization server is the one selected for the connection, and refuses to send a sign-in or a refresh anywhere else. That check is unchanged. Den no longer sets, reads or reports the mark: - A binding violation still drops the cached discovery, and the sign-in or call that hit it fails with its own error and can be retried. - Search, execute_capability, the per-connection routes, scripts, the Slack assistant, Workbot's app list and the workflow library treat the connection like any other. - The connection API no longer returns issuerReviewRequired, and the authorization server review no longer returns reviewRequired. Den web drops the "Review sign-in server" action and the notice that the sign-in server changed. Choosing an authorization server stays. - A non-admin whose sign-in needs saved credentials cleared is still told to ask an admin. The column oauth_issuer_review_required_at is left in place, unused, so instances running during a rollout keep working; a later migration drops it. The readiness fingerprint keeps a null in its place so saved fingerprints still match. The desktop app still understands the state, because older Den servers can send it. A connection that carries the old mark runs its tool again; the test for that goes through the real app. Signed-off-by: Jalil <reachjalil@gmail.com>
Contributor
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What is this about?
OAuth connections to provider MCP servers. Den no longer blocks a connection until an admin reviews its sign-in server.
What problem does it solve?
The block was meant for a provider whose sign-in server changed. It also fired on setup errors and on connections nobody had signed in to, and it never cleared, so a connection that tripped it stayed unusable.
What was the situation before?
A blocked connection showed as an error in search, refused every tool call, and offered admins a "Review sign-in server" action. Nothing else released it, even after the provider's sign-in information was consistent again.
How is this implemented?
oauth_issuer_review_required_atacross persistence, capabilities, the per-connection route, the dashboard and sign-in routes, the Slack assistant and Workbot's app list.packages/enterprise-mcp-clientstill refuses to send a sign-in or refresh to any authorization server but the selected one.issuerReviewRequired; the authorization server review dropsreviewRequired.What else changed?
nullin its place.Release note
Connections are no longer blocked until an admin reviews the sign-in server; a connection stuck in that state works again.
Evidence
ee/apps/den-api/test/api/mcp-connection-route-access.api.test.ts— before: a connection carrying the review mark refuses its tool as blocked; after: the tool runs. No E2E; the admin screens only lose a state that can no longer occur.