AI Code Review
Harness AI Code Review runs an AI agent against a pull request and reports each review criterion as a status check, on GitHub and on Harness Code Repository.
The agent does not read the diff in isolation. It calls Harness tools to bring delivery context into the review, so it can flag a change that is correct in the file and wrong for the system around it.
You define what the agent checks. Review criteria are set at a space or at a repository, and a repository inherits from the spaces above it, so a standard can be applied once and enforced everywhere.
New to AI Code Review?
Overview and key concepts
Learn what the review agent evaluates, how criteria work, and what a review produces.
Get started
Turn on AI Code Review for a GitHub or Harness Code repository, and verify the first review.
What is supported
Check supported platforms, scopes, result states, and known boundaries.
Use AI Code Review
GitHub integration
Connect a GitHub repository through a Harness connector and keep the link healthy.
Harness Code Repository
Enable reviews on a repository that already lives in Harness Code.
Define your first review criteria
Tell the agent what to check, and verify the criterion reaches a pull request.
Scope and inheritance
Apply one review standard across many repositories, and know which level wins.
What Harness creates
Every resource onboarding adds to your account, and what offboarding leaves behind.