Triage issues with an agent

Use this after you have a real projectId from get_connection or get_projects. One issue at a time. Include dashboardUrl when you tell the human what you changed.

1. Find open issues

search_errors with projectId, status: open, and environment. Use development for a local key and production for a live failure.

2. Read the issue

get_error with issueId. The result has the latest instance (stack, release, environment, lastSeenAt) and comments from earlier notes.

3. See whether it is still firing

search_events with the same projectId, environment, and issueId. Set from to the time of the suspected fix when you have one.

4. Decide

• The code no longer contains the failure and events for that issueId have stopped: update_issue_status with status: resolved and a note that says what changed.
• The issue is noise: update_issue_status with status: ignored and a note that says why.
• It is still firing, or you have not checked: leave it open. add_issue_comment with what you found.

note and body are required for those writes. update_issue_status stores the note as a comment. Both tools need mcp:write (included in install scopes) and the membership capability to update issue status.

Status rules

• open can become resolved or ignored.
• resolved and ignored can only return to open. They cannot switch to each other. Reopen with update_issue_status and status: open.
• A later event with the same fingerprint reopens a resolved issue.
• An ignored issue reopens when that project's reopen-ignored setting is on. The default is on.

Do not resolve a list of issues in one call. There is no bulk status tool.