Anthropic opens automatic code reviews to Claude Team and Enterprise
Admins can choose review depth for each repository. Anthropic says most reviews now take 6–10 minutes, down from about 20 minutes during the preview.
Loading page…
Admins can choose review depth for each repository. Anthropic says most reviews now take 6–10 minutes, down from about 20 minutes during the preview.
Listen to this story
Anthropic made Code Review generally available to Claude Team and Enterprise on October 15, 2026, moving its automated pull-request reviewer out of research preview. Organizations enable it centrally through Claude Code settings and the GitHub App, then choose repositories and review depth per repository, allowing a scoped rollout. Anthropic reports most reviews now take 6–10 minutes versus about 20 minutes in preview, but provides no numerical evidence for its claimed improvement in bug detection. A separate credit offer is limited to eligible Enterprise organizations and expires 30 days after claiming.
Administrators can assign Quick, Standard, or Deep reviews to individual repositories instead of setting one depth for the whole organization.
Eligible Claude Enterprise organizations can claim at least $10,000 in credits by January 15, 2027; credits expire 30 days after the claim.
Preview users asked for less noisy reviews, checks better matched to the change, and faster results.
Teams on Claude Team and Enterprise can now have proposed code changes reviewed automatically after an administrator enables the service. Anthropic moved Code Review out of research preview on October 15, 2026, pairing general availability with faster checks and controls that let administrators choose how deeply it reviews each code repository.
Code Review works on new pull requests—proposed changes to a codebase. The setup happens at the organization level, rather than requiring each developer to configure a review. Administrators select the repositories, or code projects, where the service should run.
Administrators can assign Quick, Standard or Deep reviews to each repository. Anthropic positions the choices as a way to give routine codebases a quick pass while reserving deeper review for the most critical ones. The setting applies by repository, not as one review level imposed across the whole organization.
Anthropic says preview teams found real issues before a human reviewer examined the pull request. But those teams also wanted quieter reviews, checks sized to the change and faster results. That feedback frames the release as an effort to make automated review easier to work with, not simply to produce more comments.
The company also runs Code Review on its own pull requests, using that experience to identify where reviews slow its engineers down. For general availability, it says most reviews finish in 6–10 minutes, compared with about 20 minutes during the preview. It describes the improvement as roughly three times faster.
The findings themselves are meant to be brief and actionable. Anthropic says they explain the problem, the conditions that trigger it and the fix in a few sentences. It also claims improved bug detection, but supplies no numerical comparison for that quality claim in the announcement. The speed figures and quality assessment are Anthropic's own.
Alongside the release, eligible Claude Enterprise organizations can claim at least $10,000 in Code Review credits by January 15, 2027. The offer is subject to terms and is specifically for eligible Enterprise organizations; general availability also includes Team customers.
The claim deadline and the spending window are separate. Credits expire 30 days after an organization claims them, so January 15 is not a promise that an earlier claim remains usable until then. Teams considering the offer need to account for that shorter window when scheduling their evaluation.
Loading discussion...
Join the conversation
Explain what would make automatic reviews helpful rather than distracting.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.