New Claude tool uses AI agents to find bugs in pull requests

Anthropic’s Claude Code Review is a new tool, available as a research preview beta for Team and Enterprise plans, that sends a team of AI agents to examine every pull request.

“We needed a reviewer we could trust on every PR. Code Review is the result: deep, multi-agent reviews that catch bugs human reviewers often miss themselves. It’s a more thorough (and more expensive) option than our existing Claude Code GitHub Action, which remains open source and available,” the company explained in a blog.

How Claude Code Review works

The system dispatches multiple agents that look for bugs in parallel. Findings go through a verification step to filter out false positives, and confirmed issues are ranked by severity.

Agents do not approve pull requests. Results appear directly on the pull request as a single overview comment, with specific bugs noted in inline comments.

Reviews scale with pull request size and complexity. Larger changes are assigned more agents and receive deeper analysis.

“We’ve been running Code Review internally for months: on large PRs (over 1,000 lines changed), 84% receive findings, averaging 7.5 issues. On small PRs under 50 lines, that drops to 31%, averaging 0.5 issues. Engineers largely agree with what it surfaces: less than 1% of findings are marked incorrect,” the company said.

Cost and availability

Code Review is billed based on token usage and scales with pull request size, complexity, and the number of issues that require verification. The average cost ranges from $15 to $25 per review.

Admins can set a total monthly spending limit across reviews, enable reviews only for selected repositories, and track reviewed pull requests, acceptance rates, and total costs.

To enable the feature, admins install the GitHub app in Claude Code settings and select repositories to review.

Once enabled, reviews run automatically on new pull requests without additional configuration.