Skip to contentSaltar al contenido
Kensa documentation

Troubleshoot Kensa

Resolve common installation, sign-in, provider, repository, connection, and automation problems with bounded next actions.

Start with the status message shown by Kensa. This diagnostic identifies the surface that failed—installation, account, provider, repository, connection, or automation—without exposing a token or internal stack trace. Fix one boundary at a time, then retry the action.

Installation, updates, and sign-in

  • The app will not open on macOS — follow the install guide's quarantine-removal step, then launch the copied application again. Do not run a copied command from an untrusted source.
  • An update does not appear or download — confirm you can reach the release source with the configured GitHub access, then restart Kensa and check again.
  • You cannot sign in — confirm the OAuth flow completed in the browser, then return to the desktop app. If the session is no longer valid, sign in again rather than reusing an old token.

Agent CLI or model problems

  • A provider is missing — install that provider's CLI, reopen the prerequisites check, and make sure the executable is on your PATH.
  • A provider is installed but not ready — complete that CLI's own authentication setup and follow the readiness hint Kensa displays.
  • A model cannot run — select a model currently available to that provider or remove the stale pin. A model choice for one CLI is not portable to another.
  • The review fails or stalls — inspect its provider and model attribution, then retry only after resolving the reported authentication, quota, network, or model condition.

Repository and connection problems

  • Deep Review is unavailable — configure a valid local clone path. Deep Review and scheduled Deep Review need a clone; Quick Review does not require the same local checkout.
  • A local diff is empty or cannot fetch — confirm the selected comparison has changes, check remote access, and use the fetch indicator to retry. See Review local changes.
  • An integration test fails — re-check the provider host, email, and credential fields the connection form requests, then re-authorize or replace the connection. Never share the token in a support message.

Automation problems

  • Run now finds nothing — preview the current PR criteria. The filters may exclude all open pull requests or skip reviews already recorded for those PRs.
  • A schedule cannot start — make sure the repository has remote PRs and that another auto-review batch is not already active.
  • A deep scheduled run is blocked — add a local clone or change the schedule to Quick Review.
  • Nothing was posted — check whether publishing is enabled and whether the selected severities matched any findings.

Next step

Return to the documentation home or revisit Configure reviews, agents, and skills.