Claude Code Review goes GA: reviews in 6 to 10 minutes, depth set per repository, and it never approves or blocks

Anthropic's announcement starts from a plain premise: coding agents keep increasing the pace at which teams write code, and the review bottleneck hasn't eased. Code Review is now out of research preview for Claude Team and Enterprise organizations, a release reported on Oct 15. The preview we covered in March ran around 20 minutes a review.

What changed for general availability

Preview users said it caught real issues before a human looked, but wanted reviews that were quieter, sized to each change, and faster. The answer on speed is that most reviews now finish in about 6 to 10 minutes, down from about 20. On sizing, admins choose Quick, Standard (the default), or Deep for each repository, so a routine codebase gets a light pass and a critical one gets a deep review; a more thorough level costs more. On noise, findings are meant to state the problem, what triggers it, and the fix in a few sentences. Anthropic also says it runs Code Review on its own pull requests.

The shape of the reviewer

The documentation describes four stages: several agents examine the diff and surrounding code in parallel, a verification step filters out false positives, the survivors are deduplicated and ranked by severity, then posted. By default it focuses on correctness, bugs that would break production, not formatting or missing test coverage, and it works from reading the diff and surrounding code, without building or running it. Two governance choices matter for teams running agents around the clock. The reviewer's instructions come from REVIEW.md on the base or default branch, never from the pull request's own changes, so an agent-authored PR cannot rewrite its own grading rules. And a review never approves or blocks the pull request, and its check run doesn't conclude as a failure. That makes it a fast first pass, not a merge gate: the docs warn that a review with no findings doesn't mean the code is error-free or secure, so the acceptance decision, and the human capacity behind it, still sets the pace. Compare CRBugBench on what review agents still miss.