Claude Code Review: Setup, Pricing and Limits in 2026

Last Updated: November 10, 2026

Summary: Claude Code Review is Anthropic's managed service that automatically reviews GitHub pull requests, using several AI agents to find bugs, verify them and post ranked inline comments. It is available only on Claude Team and Enterprise plans and is billed per review on token usage through usage credits, separate from seat prices. Anyone on any plan can instead run the /code-review command inside Claude Code, which counts toward normal plan usage.

The name "Claude Code Review" now covers four things: a GitHub service, a terminal command, a deeper cloud review and a GitHub Action. Pick the wrong one and you either overpay or get less than you expected. This guide covers each option, setup, cost, limits, how it compares to GitHub Copilot, CodeRabbit and OpenAI Codex, and when to trust its findings. Details were checked against Anthropic's official documentation and pricing page on November 9, 2026.

Claude Code Review is a managed service in which Claude reviews the pull requests in your GitHub repositories and posts its findings as inline comments. According to Anthropic's Code Review documentation, several agents read each change alongside the surrounding code, looking for logic errors, security vulnerabilities and broken edge cases, and a verification step checks each finding before it is reported.

Anthropic announced Code Review on March 9, 2026, as a research preview for Team and Enterprise customers, describing it in the launch announcement as "built for depth, not speed." A review never approves or blocks a pull request, so the merge decision stays with your team.

By default, Claude Code Review focuses on correctness: bugs that would break production, not formatting preferences or missing test coverage.

A Claude Code Review runs on Anthropic's infrastructure in four stages. Several agents examine the diff and surrounding code in parallel (find), a verification step filters out false positives (verify), remaining findings are deduplicated and ranked by severity (rank), and the review posts inline comments, a summary and a check run named "Claude Code Review" (post).

A Claude Code Review reads code rather than running it: Anthropic says it works "without building or running your code," has read-only repository access and never pushes commits. Anthropic chooses the models, and admins get no model choice.

Each finding gets one of three severity markers, per Anthropic's guide to reading reviews:

The Claude Code Review check run never concludes as a failure, so it will not fail a required check. Teams that want a hard gate can read the severity counts with the GitHub CLI and fail their own CI job when Important findings are above zero, using a script Anthropic documents.

Anthropic offers five ways to get Claude to review code, per its Code Review overview, the /code-review guide and the ultrareview guide.

Our rule of thumb: individuals start with /code-review, Team and Enterprise organizations wanting hands-off coverage turn on the managed service, and teams on Bedrock, Google Cloud or Foundry, or who want full workflow control, use the GitHub Action.

An Owner of your Claude organization turns on Code Review in Admin settings on claude.ai, installs the Claude GitHub App and picks repositories. No workflow files are needed because reviews run on Anthropic's infrastructure. Anthropic's setup guide lists these steps:

Watch one default: each new repository starts at "After every push," which runs the most reviews. The alternatives are "Once after PR creation" and "Manual." Owners also pick a review level: Quick, Standard (default) or Deep. Developers with write access can comment @claude review for one review, which also works on drafts and forks that automatic reviews skip, or @claude review always to review every later push.

Claude Code Review has no flat per-review price. Anthropic's costs documentation says reviews are billed on token usage through usage credits, separately from plan usage, and charged to the organization. Cost scales with review level, change size, code complexity and how many candidate issues need verification.

At launch, Anthropic said reviews "generally" averaged $15 to $25 each and took about 20 minutes on average in its testing. Treat those as launch-era figures; the current docs give no average and ask teams to measure their own.

You also need the right plan first. Claude's live pricing page listed these business plans on November 9, 2026:

Anthropic's estimate formula is average cost per review, times pull requests per month, times pushes per pull request if you review every push. As an illustration at the $20 launch-era midpoint: a team that opens 100 pull requests a month on "Once after PR creation" would spend about $2,000 a month (100 x $20). Switch the same team to "After every push" with an average of three pushes per pull request, and the estimate triples to about $6,000 (100 x 3 x $20). Real push costs may come in lower, since Anthropic says later push-triggered reviews focus on what changed since the last review.

The review trigger often matters more than the review level. Anthropic suggests a pilot: set a spend limit, turn on a few representative repositories at "Once after PR creation" or "Manual," and read each repository's Average cost column before rolling out widely. Once the spend limit or credit balance runs out, reviews are skipped with an explanatory comment on the pull request.

The /code-review command (alias /review) reviews your current branch inside a Claude Code session, on any plan, and counts toward normal usage. Anthropic's /code-review guide says it runs as a background subagent with its own context window.

You can pass a target (pull request number, branch, path or commit range), --fix to apply findings, or --comment to post them on a GitHub pull request or GitLab merge request. At low effort it reports only its most confident findings. If you hit plan caps often, our guide to Claude usage limits explains how they work.

Ultrareview is a deeper review that runs a fleet of reviewer agents in a cloud sandbox, started with /code-review ultra. Anthropic's ultrareview guide calls it a research preview that independently verifies every finding. Pro and Max accounts get three free runs, once. After that, a review typically costs $5 to $25 in usage credits and takes 5 to 10 minutes.

Teams tune Claude Code Review with two files committed to the repository: CLAUDE.md for project-wide rules and REVIEW.md for review-only rules. Anthropic's customization guide suggests using REVIEW.md to:

Claude Code Review reads REVIEW.md from the base or default branch, never from the pull request itself, so a contributor cannot rewrite the rules in their own pull request. Rules that lower severity, exclude paths or cap nits apply only when the default branch is protected. The local /code-review command ignores REVIEW.md, so shared rules belong in CLAUDE.md. Our Claude Code best practices guide covers writing a good CLAUDE.md.

The Claude Code GitHub Action, anthropics/claude-code-action@v1, runs Claude inside your own GitHub workflows. The fastest setup, per Anthropic's GitHub Actions documentation, is to run /install-github-app in Claude Code, which installs the app, stores your API key or subscription token as a repository secret and opens a pull request with the workflow files.

Anthropic's example review workflow runs the official code-review plugin on each pull request and posts inline comments. The Action works with the Claude API, Claude subscriptions, Amazon Bedrock, Google Cloud and Microsoft Foundry, and each run uses API tokens (or your subscription) plus GitHub Actions minutes. Our Claude Code plugins guide explains how plugins like code-review are installed.

Claude Code Review costs more per review than GitHub Copilot code review but is built for depth, while CodeRabbit charges per developer seat instead of per review. The table compares what each vendor's official documentation says as of November 9, 2026.

Sources: GitHub's Copilot code review docs, CodeRabbit's plans page and OpenAI's Codex GitHub guide. GitHub says Copilot reads instructions from the pull request's head branch, while Claude Code Review reads REVIEW.md only from the base or default branch, which matters when outside contributors open pull requests. OpenAI says Codex flags only P0 and P1 issues on GitHub.

Which should you pick? Claude Team or Enterprise customers get the tightest fit from Claude Code Review, especially on critical repositories where a missed bug costs far more than a review. Teams wanting many cheap reviews will find Copilot's per-review cost lower. See our explainer on what GitHub Copilot is and our OpenAI Codex pricing guide for those plans.

Anthropic reports strong internal results, but its documentation warns that Claude Code Review can miss bugs and that a clean review does not mean code is error-free or secure. The launch post reported, from Anthropic's own teams:

Those are vendor-reported numbers, not an independent benchmark.

Sameer's take: the biggest business mistake with AI is trusting output without checking it. That applies here too. Treat Claude Code Review as a strong second reader, not a replacement for a human approver.

The managed Claude Code Review service works only with GitHub (including GitHub Enterprise Server), not GitLab or Bitbucket. Anthropic says it is unavailable to organizations with Zero Data Retention or the HIPAA configuration, is not covered by its Business Associate Agreement and does not yet support customer-managed encryption keys. Bedrock, Google Cloud and Foundry users still need a Claude Team or Enterprise organization.

What is Claude Code Review?
Claude Code Review is Anthropic's managed service that reviews GitHub pull requests with several AI agents, verifies findings and posts them as ranked inline comments. It never approves or blocks a pull request.

How much does Claude Code Review cost?
Claude Code Review has no flat price. It is billed on token usage through usage credits, separately from seat prices, and Anthropic's March 2026 launch post said reviews generally averaged $15 to $25 each depending on pull request size and complexity.

Is Claude Code Review included in the Pro or Max plan?
No. The managed Code Review service is available only on Claude Team and Enterprise plans. Pro and Max users can run /code-review in Claude Code and get three free ultrareview runs.

How do I trigger a Claude code review on a pull request?
Depending on the repository setting, reviews run when a pull request opens, after every push, or on request. Anyone with write access can comment @claude review to start one.

Is Claude Code Review better than GitHub Copilot code review?
Claude Code Review is built for depth with multi-agent verification but costs more per review. GitHub estimates Copilot reviews at about $0.05 to $5 in AI credits depending on effort, so Copilot suits high-volume review while Claude suits critical changes.

Claude Code Review suits Team and Enterprise organizations that want a deep, verified second look at every pull request without maintaining workflow files. Start on "Once after PR creation" or "Manual," set a spend limit and measure your own average before going wide. Individuals get most of the value from /code-review. Either way, keep a human as the final approver.

⭐ Get more AI Business Weekly in Google: add us as a preferred source on Google.

Written by Sameer Khan with AI assistance, fact-checked against the sources linked above.