Anthropic makes Claude Code Review generally available for Team and Enterprise

What Anthropic made generally available

Anthropic published Code Review is now generally available on the Claude product site, stating that Code Review is out of research preview and available to Claude Team and Enterprise organizations. The post frames the product as a response to a familiar bottleneck: coding agents increase how fast teams write code, while human review queues do not shrink at the same pace.

The GA note highlights three product changes since preview. Speed: most reviews now finish about three times faster, from roughly 20 minutes down to 6–10 minutes. Quality: Anthropic says reviews catch more production-relevant bugs and present shorter findings that state the problem, what triggers it and the fix. Flexibility: admins choose Quick, Standard or Deep for each repository so routine codebases get a lighter pass and critical ones get a deeper review.

Setup instructions sit in the Code Review documentation and the setup guide. Owners enable the feature under Admin settings → Claude Code, connect GitHub, install the Claude GitHub App, select repositories and finish the switch. Reviews run on Anthropic infrastructure; teams do not add workflow files to the repo for the managed service.

How a review actually runs on a pull request

By default Code Review focuses on correctness—bugs that would break production—rather than formatting preferences or missing tests. Teams can steer what gets flagged with REVIEW.md and CLAUDE.md files in the repository. Instructions are read from the base or default branch, never from the pull request's own changes, which blocks a PR from rewriting the reviewer's brief.

A review runs in four stages on Anthropic infrastructure: several agents examine the diff and surrounding code in parallel; a verification step checks candidate findings against the code and filters false positives; remaining findings are deduplicated and ranked by severity; results are posted on the pull request. Each review leaves inline comments, a summary in the review body and a check run named Claude Code Review that does not conclude as a failure. When nothing is found, the check title reads that no issues were found and Claude posts the same short note on the PR.

Access inside a review is read-only on the repository: the service does not push commits. It works from reading the diff and surrounding code without building or running the project. Anthropic chooses the Claude models used; admin settings do not offer a model picker. Depth is controlled through the review level instead.

That design sits next to other automated review launches this month, including GitHub's Agentic Autofix and AI Scan general availability for code scanning. GitHub's tools target vulnerability fixing in the scanning pipeline; Anthropic's Code Review targets pull-request correctness comments. They can coexist on the same repository.

Quick, Standard and Deep review levels

Review level trades thoroughness for time and spend. Quick is the lightest, fastest and cheapest pass. Standard is the default balance of depth, speed and cost. Deep is the most thorough, slowest and most expensive. An Owner sets an organization default and can override it per repository—for example Quick on a routine service and Deep on a payments codebase.

A new level applies to reviews that start after the change. Anthropic's cost docs (linked from the product pages) explain that level is only one driver of spend; how often reviews run and how large each diff is also matter. AI Lookout has not independently timed the 6–10 minute GA claim; that figure is Anthropic's.

Who can turn it on and which plans qualify

Availability is limited to Team and Enterprise plans. Code hosts include GitHub.com, GitHub Enterprise Server and GitHub Enterprise Cloud with data residency (Owners enter the enterprise ghe.com hostname for residency setups). Organizations need a Claude org on those plans even if developers reach Claude Code through Amazon Bedrock, Google Cloud's Agent Platform or Microsoft Foundry.

Several configurations are excluded. Code Review is not available with Zero Data Retention enabled, with Anthropic's HIPAA configuration applied, or under Anthropic's Business Associate Agreement. It is also not yet available for organizations that use customer-managed encryption keys. Those limits matter for regulated customers who otherwise already buy Enterprise seats.

Turning the product on requires an Owner role in the Claude organization, a GitHub organization owner to install and link the Claude GitHub App if it is not already linked, and write access to each repository added from the connected GitHub account. The setup dialog can cover repositories one by one or, on github.com with a spend limit set, an entire GitHub organization at once—an option Anthropic warns can raise spend significantly.

When reviews run, and how developers request one

Each repository has a Review runs setting. Once after PR creation reviews when a pull request is opened, reopened or marked ready. After every push reviews on branch updates, catching new issues as the PR evolves and auto-resolving threads when flagged issues are fixed. Manual (@claude review) starts nothing automatically; someone with write access comments to request a review, or uses @claude review always to subscribe that PR to subsequent pushes.

Automatic reviews skip drafts and pull requests from forks regardless of the setting; a maintainer can still request those manually with the comment command. New repositories added from the dialog start at After every push, which runs the most reviews. Tip comments—short notes explaining why a review did not start—are on by default and can be disabled.

Developers without the Owner role cannot change org-level depth or repository coverage, but they can request reviews, subscribe a PR to push reviews and tune flagging with REVIEW.md and CLAUDE.md. That split mirrors how Claude Code Projects opened to waitlisted Pro and Max users: admins gate the capability, developers drive day-to-day use.

Credits, cost controls and what still is not covered

The GA post includes a credit offer: eligible Claude Enterprise organizations can claim at least $10,000 in Code Review credits by 15 January 2027. Credits expire 30 days after they are claimed, and terms apply. Anthropic does not publish a public price card for each Quick/Standard/Deep review in the GA article; spend limits and usage views live in the cost docs linked from the admin pages.

Important non-goals stay explicit. A review with no findings does not mean the code is secure or bug-free. Code Review uses AI and may miss bugs or vulnerabilities. The check run never blocks merge; humans keep the merge decision. The Claude GitHub App's permission set is broader than a single review uses because it is shared with other Claude features.

How this compares with other Claude review options

Anthropic's docs compare several review paths. Managed Code Review is the Team/Enterprise GitHub service described here. Anyone on any plan can still run /code-review inside a local Claude Code session. Ultrareview runs from Claude Code in a cloud sandbox for signed-in claude.ai users. Teams can also wire Claude into their own GitHub Actions or GitLab CI with an API key, or write a custom reviewer subagent. Claude Opus 5.5 fast mode leaving research preview is a separate speed product for the same model family; it does not replace Code Review.

For organizations already standardising on Claude Code for authoring, the managed PR path removes the need to keep a custom Actions reviewer alive. For organizations that need ZDR, HIPAA or CMEK, the comparison table effectively points back to self-hosted CI or local /code-review until Anthropic extends the managed service.

What to watch after GA

Worth watching after GA: whether Anthropic publishes clearer per-review pricing; whether support arrives for ZDR/HIPAA/CMEK customers; how REVIEW.md patterns evolve in public repos; and whether independent teams measure false-positive rates against the preview baseline. Also watch whether the $10,000 Enterprise credit offer changes adoption curves before the 15 January 2027 claim deadline.

Until those land, the actionable fact is narrower: Team and Enterprise Owners can turn on a managed, non-blocking GitHub review service that Anthropic says is roughly three times faster than its research preview, with per-repo depth controls and an Enterprise credit window.