--- title: "Claude Code Review GA: Team/Enterprise Setup and Review Tiers" description: "Claude Code Review left research preview on Oct 15, 2026 for Team and Enterprise. Setup steps, how to choose Quick, Standard or Deep, and a rollout checklist." url: https://cdnping.org/en/blog/claude-code-review-ga language: en published: 2026-10-19 category: "Developer tools" tags: ["Claude Code","code review","GitHub","Team","Enterprise"] source: cdnping.org license: Free to quote and summarize with a link to the url above. --- # Claude Code Review GA: Team/Enterprise Setup and Review Tiers Claude Code Review left research preview on October 15, 2026 and is now generally available to Claude Team and Enterprise organizations. Setup is a one-time job for an organization owner: enable it in Claude Code settings, install the GitHub App, pick your repositories, and new PRs are reviewed automatically. If you are on an individual subscription, you can use the CLI's `/code-review` to review a local branch instead. This article draws on [Anthropic's GA announcement](https://claude.com/resources/articles/code-review-is-now-generally-available), the [Code Review docs](https://code.claude.com/docs/en/code-review) and the [changelog](https://code.claude.com/docs/en/changelog), as of October 2026. Menu names, limits and billing details change between versions, so treat the official docs as the source of truth. ## What changed with GA? The announcement lists these changes: - The research preview is over, and both Team and Enterprise organizations can use it. - Anthropic says most reviews are about 3x faster than in the preview, dropping from roughly 20 minutes to 6 to 10 minutes. - The tier can be set per repository: Quick, Standard (the default) and Deep. - Eligible Enterprise organizations can claim at least $10,000 in Code Review promotional credit before January 15, 2027. Claimed credit expires after 30 days. Check the official terms for details. I did not see a per-review price in the announcement, so refer to the official docs for cost. ## How does a review work? According to the docs, a review runs on Anthropic's infrastructure in four steps: several agents read the diff and the surrounding code in parallel; a verification step filters out false positives against the code; the remaining findings are deduplicated and ranked by severity; and the results are posted back to the PR. Each review leaves the following on the PR: - Inline comments on specific lines of code, plus a summary. - A check run named Claude Code Review. Its conclusion is never "failed." If nothing is found, the title says no issues were found. Two things to keep in mind. First, a review neither approves nor blocks a merge; your team still decides. Second, it only reads the diff and surrounding code and does not build or run your code. Anthropic also warns that AI review can miss bugs, so "nothing found" does not mean the code is fine. For the default focus areas and what you can tune, see the official docs. ## How do you set it up for Team or Enterprise? 1. Check eligibility: you need a Team or Enterprise plan, and your code must be hosted on GitHub. If your organization has special data retention, compliance or encryption-key settings, check the docs first to confirm support. 2. An admin enables Code Review in Claude Code settings. 3. Install the GitHub App and select the repositories to review. 4. Choose a review tier (Standard by default). You can set an organization default or override it per repository. 5. From then on, new PRs trigger reviews according to each repository's settings (once when the PR is opened, after every push, or only on request). Option names are in the official docs. Once it is on, you can start a review manually by commenting `@claude review` on a PR. This requires write access to the repository. For how draft PRs and fork PRs are handled, see the official docs. ## Quick, Standard or Deep? Anthropic only describes the tiers relative to each other: Quick is the fastest and cheapest; Standard balances depth, speed and cost and is the default; Deep is the most thorough, and also the slowest and most expensive. I did not find per-tier timing or pricing in the announcement. The table below is a way to think about it, not official guidance: | Repository type | Tier to try first | Why | | --- | --- | --- | | Docs, scripts, small internal tools | Quick | Mistakes are cheap; you want speed | | Everyday product code | Standard | It's the default; observe for a week or two | | Payments, permissions, data migrations and other critical paths | Deep | A missed bug costs more than the extra review spend | I would run a batch of PRs on Standard, compare against the issues your human reviewers caught, and only then move repositories up or down a tier. For where to set cost and spend limits, see the official docs. The changelog also notes that a review that hits its time limit without verifying any findings is marked incomplete and is not billed. ## How do you tune reviews to your team's conventions? You can put a `REVIEW.md` (alongside your `CLAUDE.md`) in the repository to adjust what the review focuses on; see the official docs for the format. According to the changelog, the check run now shows a notice when `REVIEW.md` did not take effect, so look there first if your rules seem ignored. The CLI's `/code-review` also gained a `--max-findings` flag to cap the number of reported findings. ## Pre-rollout checklist - [ ] Confirm your organization's compliance settings are not in the officially listed unsupported cases - [ ] The target repositories are on GitHub and the App has access to them - [ ] Pick a tier for each repository - [ ] Confirm cost and spend-limit settings - [ ] Commit a `REVIEW.md` and check whether the check run warns that it was not applied - [ ] Agree as a team that review comments are advisory and merges still need human review ## What if a review doesn't trigger or the result looks wrong? Start with the notice on the PR's check run, then the troubleshooting section of the official docs. If you suspect a problem with the service itself, see cdnping's [/en/down/claude](/en/down/claude) and [/en/status/anthropic](/en/status/anthropic). For connection problems with Claude Code on your machine, see the [four-layer network checklist](/en/blog/claude-code-connection-troubleshooting). For retries and usage-limit errors, see [529 Overloaded and usage limits](/en/blog/claude-code-529-overloaded-retry-idle-compaction); for commands blocked by auto mode, see [auto mode troubleshooting](/en/blog/claude-code-auto-mode-default). If you want to configure a custom endpoint in your own tooling, see [custom API configuration](/en/blog/configure-custom-api-endpoint-claude-code-codex). ## What about individual developers or teams without a Team plan? You can run `/code-review` in a Claude Code session to review a local branch, or wire Claude into your own CI using the methods in the official docs. Trigger method, where it runs and billing differ between these options, so check the official docs. ## Sources - [Code Review is now generally available (Anthropic, 2026-10-15)](https://claude.com/resources/articles/code-review-is-now-generally-available) - [Code Review docs](https://code.claude.com/docs/en/code-review) - [Claude Code changelog](https://code.claude.com/docs/en/changelog)