AI Code Review

Access Harness AI Code Review documentation

Harness AI Code Review runs an AI agent against every pull request and reports what it finds as status checks and comments. It works on GitHub repositories and on Harness Code repositories. Review standards are defined as criteria and inherited down your account hierarchy, governed by the same role-based access control as the rest of the platform, and evaluated by your own model provider. Because the agent runs inside a Harness pipeline, it reaches Harness delivery data that a diff-only reviewer cannot.

AI Code Review onboarding illustration

Get Started

Overview & Key Concepts

What a review criterion is, where settings live, and what a review run produces.

Get Started

Take one repository from nothing to a reviewed pull request.

What Is Supported

Platforms, states, and the boundaries of the beta.

Platforms

Harness Code

Reviewed in place, with no connector and no import to wait for.

GitHub

Linked through a Harness connector, with results reported back onto the pull request.

Configure

Settings Reference

Every field on the AI Code Review setting, and what each one changes.

Scope & Inheritance

Apply one review standard across many repositories.

What Harness Creates

The service account, role, token, pipeline, and triggers onboarding leaves behind.

Connectors & Models

Bring your own key, or use the Harness AI gateway.

Workflows

First Review Criteria

Tell the agent what to check, starting with two criteria rather than twenty.

Review a Pull Request

Follow a single change through a full review, end to end.

Criteria Best Practices

Reduce noise before it costs you the team's attention.

Resources

API Reference

Every onboarding and settings endpoint.

Permissions & RBAC

What each action requires, and at which scope it is evaluated.

Limits & Quotas

What is bounded, and what only looks unbounded.

FAQs

Common questions and the failures teams hit most often.