Flaky-test detection needs a history of runs, and the setup step — getting results uploaded from your CI — is where most teams stall. Shipping that as a tool means the agent can do the onboarding as well as the fixing, in the same conversation. On the analysis itself: root cause matters more than the fix suggestion, because flakiness is usually a shared-state or timing problem that a naive retry hides rather than solves. Read the cause before applying the patch.
Trunk detects flaky tests across CI runs. The endpoint hands that analysis to a coding agent instead of a dashboard.
- Fix-flaky-test — the root-cause analysis for a specific flaky test, with a suggested fix
- Setup-trunk-uploads — generate the plan for uploading your test results in the first place
OAuth with OpenID Connect against your Trunk account. A single command adds it in Claude Code; other clients take the URL.
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen