Anthropic made Code Review for Claude Code generally available on October 15, 2026, ending its research preview for Claude Team and Enterprise organizations. An administrator can connect the Claude GitHub App, choose repositories and a trigger, and have parallel agents review pull requests automatically. Anthropic says reviews now usually finish in 6–10 minutes, down from about 20 minutes during the preview.
What general availability changes
Code Review is now a supported Claude Code feature for eligible Claude Team and Enterprise organizations rather than a research-preview experiment. Organization owners or primary owners enable it in Claude Code settings, install the Claude GitHub App, and select the repositories to monitor.
As an Amazon Associate I earn from qualifying purchases.
After setup, the service analyzes pull-request changes and surrounding repository code, then posts a summary and inline comments. Anthropic says the default review focuses on correctness problems that can affect production: logic errors, security vulnerabilities, broken edge cases and regressions.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Administrators can set review depth independently for each repository:
#1 Best Overall
| Review depth | What is established |
|---|---|
| Quick | A faster, lighter review option; Anthropic does not publish a fixed time or issue-count guarantee. |
| Standard | The default depth for a repository. |
| Deep | A more extensive review option; Anthropic does not publish a fixed time or issue-count guarantee. |
How Claude Code Review analyzes a pull request
The system dispatches multiple specialized agents to inspect the pull-request diff and relevant parts of the codebase in parallel. A separate verification step checks candidate findings against actual code behavior. Findings are then deduplicated, ranked by severity and posted where developers can act on them: in a review summary and as inline comments on affected lines.
This is still an advisory review. Code Review does not approve pull requests and does not block merges. Teams therefore keep their existing human approval, branch-protection and CI rules.
Repository-specific guidance
Teams can add a CLAUDE.md file and a root-level REVIEW.md file to describe project conventions or additional checks. Those instructions supplement the default correctness review; they do not replace it.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Set up automatic reviews
Owners or primary owners need permission to install GitHub Apps. The documented setup sequence is:
- Open Claude Code settings for the Claude Team or Enterprise organization and enable Code Review.
- Install the Claude GitHub App, authorizing it with an account that can install GitHub Apps.
- Select the repositories that Code Review may access.
- Choose one repository trigger: review once after pull-request creation, review after every push, or wait for a manual top-level
@claude reviewcomment. - Choose Quick, Standard or Deep review depth for each selected repository.
- Add optional repository guidance in
CLAUDE.mdand a rootREVIEW.mdif the project needs checks beyond the default correctness focus.
Draft pull requests are not reviewed. A pull request must become eligible under the selected trigger before a review is dispatched.
Triggers, GitHub coverage and important exclusions
| Area | Supported behavior or limitation |
|---|---|
| New pull requests | Can trigger one review after creation. |
| Subsequent pushes | Can trigger a review after every push when that option is selected. |
| Manual review | Start one with a top-level @claude review comment. |
| Draft pull requests | Not reviewed. |
| Standard GitHub | Supported on repositories hosted at github.com. |
| Self-hosted GitHub Enterprise Server | Supported through a separate setup flow. |
| GitHub Enterprise Cloud data-residency tenants | Repositories on *.ghe.com are not supported according to the setup guide. |
| Merge control | Code Review neither approves nor blocks a pull request. |
Price, credits and review speed
Code Review is billed separately from the organization’s normal Claude usage through usage credits. Anthropic’s setup documentation gives an average of $15–$25 per review (Anthropic, 2026). That is an average, not a fixed price: the actual charge varies with pull-request size, codebase complexity and the amount of verification work required.
Eligible Claude Enterprise organizations can claim at least $10,000 in Code Review credits by January 15, 2027. Credits expire 30 days after they are claimed, so an organization should claim them close to the period in which it expects to use them and confirm its eligibility in the current administrator documentation.
Anthropic reports that most reviews now complete in roughly 6–10 minutes, compared with about 20 minutes during the preview. The published timing is a typical range, not a service-level guarantee; larger or more complex changes can take longer.
Best Value
What Anthropic has reported from the preview
Anthropic’s preview announcement reported the following internal results:
- Code output per Anthropic engineer increased by 200%.
- Pull requests receiving substantive review comments rose from 16% before Code Review adoption to 54% afterward.
- Among pull requests with more than 1,000 changed lines, 84% had findings, compared with 31% of pull requests under 50 changed lines.
- Large pull requests averaged 7.5 issues, while small pull requests averaged 0.5 issues.
- Less than 1% of findings were marked incorrect.
These figures are Anthropic’s own internal reports, not an independent benchmark or independently verified customer study. They indicate how the system performed in Anthropic’s environment, but they do not predict the exact finding rate, latency or productivity change for every repository.
When automatic review is a good fit
Review-after-creation is useful when a team wants an initial pass without waiting for a human reviewer. Review-after-every-push provides a fresh pass as a branch changes, but it can consume more credits on active pull requests. Manual review is the most controlled option for repositories where cost, timing or sensitive changes require an explicit decision.
Because the tool is advisory, it works best alongside tests, static analysis, security scanning and human review. A finding should be reproduced and evaluated in context before code is changed; an apparently valid comment can still be inappropriate for a project’s architecture or threat model.
Quick Recap
Practical rollout checklist
- Confirm that the organization is on Claude Team or Claude Enterprise and that the administrator has GitHub App installation permission.
- Start with a small set of representative repositories rather than enabling every repository at once.
- Use Standard depth initially, then move repositories to Quick or Deep according to review latency and issue complexity.
- Decide whether every push justifies another paid review or whether creation-time and manual triggers are sufficient.
- Write project-specific conventions in
CLAUDE.mdand review requirements in the rootREVIEW.md. - Keep human approval and merge-protection rules enabled because Code Review cannot enforce them.
- Monitor credit consumption, especially for large pull requests and repositories with frequent pushes.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
