AI Review
Introduction
Run your first AI Review on a connected GitHub pull request and understand the result.
Angada reviews pull request changes and reports findings with code evidence. This guide takes you from an eligible pull request to its review result.
Before you start
- The repository is connected to Angada through the GitHub App, and the installation has access to it. See GitHub integration or follow the quickstart to connect GitHub.
- Open a pull request in that repository. Draft pull requests are skipped by default, and workspace review eligibility and limits can affect whether a review runs. See Limits and caps.
Follow the first review
- Open a pull request in the connected repository, or mark an existing draft ready for review. Angada can start a review for either event when the pull request is eligible.
- Open the pull request's Checks area on GitHub. The Angada check shows that the review is Reviewing… while it runs.
- When the check finishes, read the review status and findings. Findings can include a claim, severity, confidence, and file and line evidence. Use Understand a review to interpret them.
An Angada pull request check with a review finding.Show the Angada check beside a representative review finding on a pull request.Capture pending · 1600 × 1000px
Angada's review status is separate from GitHub's review decision. It does not approve or request changes on the pull request.
| Result | What it tells you |
|---|---|
| Reviewed | The review completed with a clear verdict. |
| Reviewed (degraded quality) | The review completed but calls for a closer human look. |
| Review failed | The review could not complete. See Troubleshoot a review. |
| Skipped | A trigger, setting, or eligibility rule prevented the review. See Limits and caps. |
Key features
- Understand a review — status, verdicts, findings, and code evidence.
- Re-runs and continuity — reviews after pull request updates and how findings carry forward.
- Review settings and Limits and caps — review controls, eligibility, and defaults.
- Coding agents — hand off a finding as a prompt for a coding agent.