Claude Code Review in 2026: Pricing, Setup, Limits, and Every Claude Code Code Review Option

"Claude Code Review" can mean three things: Anthropic's managed pull request reviewer for GitHub, the /code-review command inside a Claude Code session, or a review job you build yourself on the Claude Code GitHub Action. This guide covers all three, starting with the managed service, which is what most people mean.

The managed Claude Code Review service is a multi-agent reviewer: several agents read each change alongside the surrounding code, a verification step filters out false positives, and the findings land as inline comments ranked by severity, with a summary and a check run on the pull request. It launched as a research preview on March 9, 2026 and became generally available on October 15, 2026 for Claude Team and Enterprise plans. It is billed on token usage (Anthropic's launch announcement put the average at $15 to $25 per review), and by design it never approves or blocks a pull request. The merge decision stays with your team.

That last point is a deliberate design choice, not a knock. But together with plan gating and per-review cost, it shapes whether Claude Code Review is the right reviewer for you. Below: what each option costs, who can use it, how long a review takes, how to set it up on GitHub, where the limits are, and where Macroscope fits at an average of 55 cents per review on Budget and 76 cents on Balanced. For the wider picture of Claude Code on GitHub, including the Claude Code GitHub Action and @claude mentions, see our Claude Code GitHub guide.

The 30-second summary of Claude Code Review

• What: Anthropic's managed, multi-agent AI code review for GitHub pull requests. Research preview from March 9, 2026, generally available since October 15, 2026.
• How: Parallel agents find issues, a verification step filters false positives, and Claude posts inline comments ranked by severity plus a check run named Claude Code Review.
• The price tag: Billed on token usage through usage credits, with no flat per-review price. At launch Anthropic said reviews "generally average $15-25," scaling with PR size and complexity.
• Review time: Anthropic's GA announcement says most reviews now finish in about 6 to 10 minutes, down from about 20 minutes during the preview.
• Who can use it: Team and Enterprise organizations, once an Owner turns it on. On any plan, /code-review runs a review in your own Claude Code session instead.
• What it won't do: It never approves or blocks a pull request, its check run does not conclude as a failure, and it does not push fixes.
• The usage-based alternative: Macroscope reviews pull requests at $0.05 per KB of diff (averaging 55 cents on Budget and 76 cents on Balanced), with $100 in usage credit for new workspaces, and it can also auto-approve safe PRs, fail a check on a rule violation, and open fix PRs.

What Is Claude Code Review?

Claude Code Review is a managed pull request review service from Anthropic that reviews GitHub pull requests on Anthropic's infrastructure, without a workflow file in your repository. Instead of a single model reading the diff once, Anthropic's documentation describes several agents examining the change and the surrounding code in parallel, a verification step that checks candidate findings against the code, and a ranking step that deduplicates findings and sorts them by severity before anything is posted.

It is worth separating the things people mean by Claude Code code review:

1. Claude Code Review (the managed service). The automated GitHub PR reviewer described above, for Team and Enterprise organizations, including GitHub Enterprise Server. This is the main subject of this guide.
2. /code-review in Claude Code. A command you type in a Claude Code session. Anthropic's local review docs say it works on any plan, reviews the changes on your branch for correctness bugs, and can post findings to a pull request with --comment.
3. A review workflow on the Claude Code GitHub Action. You can run Claude Code in your own CI with anthropics/claude-code-action and have it review pull requests. Anthropic's launch post described Code Review as optimized for depth and more expensive than this lighter-weight option.

All three are useful, and they solve different problems. Automated, every-PR GitHub code review is the job most teams are hiring for, and that is the lens this guide uses.

How Does Claude Code Review Work?

Claude Code Review works in four stages: find, verify, rank and post. According to Anthropic's docs, a review runs on Anthropic infrastructure like this:

1. Find. Several agents examine the diff and the surrounding code in parallel.
2. Verify. A verification step checks the candidate findings against the code and filters out false positives.
3. Rank. The findings that remain are deduplicated and ranked by severity.
4. Post. Claude posts inline comments on the lines concerned, a summary in the review body, and a check run named Claude Code Review.

By default it focuses on correctness, meaning bugs that would break production, rather than formatting preferences or missing test coverage. A review reads the diff and surrounding code without building or running your code, and inside a review its access to the repository is read-only. When it finds nothing, the check run's title reads "Code review found no issues."

On timing, Anthropic's launch announcement put the average review at around 20 minutes during the preview. The GA announcement says most reviews now finish about 3x faster, in roughly 6 to 10 minutes. That suits asynchronous review; for a quick signal on a small change before you push, /code-review in your own session is the faster loop.

This multi-agent, verify-before-reporting design is where the whole category is heading, and it matters more now that so many pull requests are themselves written by coding agents.

How Much Does Claude Code Review Cost?

Claude Code Review has no flat per-review price: it is billed on token usage through usage credits, separately from your plan's included usage, and Anthropic's March 2026 launch post said reviews "generally average $15-25." Anthropic's costs page says the cost of each review scales with the review level, the size of the change, the complexity of the surrounding code, and how many candidate issues need verification.

The number of reviews matters as much as the cost of each one. Each repository's review setting decides how often reviews run, and a repository added from the admin page starts at "After every push," which multiplies cost by the number of pushes to each pull request. Switching a repository to "Once after PR creation" or "Manual (@claude review)" runs fewer reviews.

Cost control is real. An organization can set a monthly spend limit for the Code Review service in its usage settings on claude.ai, and reviews are skipped, with a comment on the PR explaining why, when that limit, the organization's overall limit, or its usage credit balance runs out. Anthropic also recommends a small pilot: turn it on for a few representative repositories, read the average cost per review in the admin settings, and multiply by your monthly PR volume.

For comparison, a usage-based AI code reviewer like Macroscope prices code review at $0.05 per KB of diff reviewed, with a 10 KB minimum ($0.50 floor). Reviews average 55 cents on Budget and 76 cents on Balanced. Against the low end of Anthropic's launch average, that is $15 / $0.76 = about 20 times the Balanced average per review; at the high end, $25 / $0.76 = about 33 times. Macroscope's spend controls include caps per review and per PR on automatic reviews (by default $10 and $50), monthly budget limits, and turning off auto-review by repo or label, and every new workspace starts with $100 in usage credit. The difference per review is worth modeling against your actual PR volume and push frequency.

Who Can Use Claude Code Review?

Claude Code Review is available to organizations on Claude Team and Enterprise plans, and an Owner of the Claude organization has to turn it on. It is not part of individual Pro or Max plans. It works with GitHub repositories, including GitHub Enterprise Server and GitHub Enterprise Cloud with data residency. Anthropic's docs also say it is not available to organizations with Zero Data Retention enabled or with the HIPAA configuration applied, and not yet available to organizations that use customer-managed encryption keys.

That does not mean individual developers are shut out of Claude Code code review altogether. On any plan, you can run /code-review in your own Claude Code session to review your branch, point it at a pull request number, apply fixes with --fix, or post findings to the PR with --comment. What you do not get without Team or Enterprise is the managed service that reviews every pull request automatically.

How Do I Set Up and Customize Claude Code Review?

An Owner turns on Claude Code Review in Admin settings, Claude Code, on claude.ai, installs the Claude GitHub App, chooses repositories, and sets a review level and when reviews run. Reviews run on Anthropic's infrastructure, so there is no workflow file to add. Per Anthropic's setup docs:

1. Allow code review. Flip the "Allow code review" switch in the Code review section of the admin settings.
2. Connect GitHub and install the Claude GitHub App on the GitHub organization that owns the repositories, choosing which repositories it can access.
3. Choose repositories and finish setup.
4. Verify by opening a non-draft test pull request. A check run named Claude Code Review should appear within a few minutes.

Then tune it:

• Review level. Quick (lighter and cheapest), Standard (the default), or Deep (most thorough and most expensive), set for the organization and overridable per repository.
• When reviews run. Once after PR creation, after every push, or manual. Anyone with write access can comment @claude review to request a single review, or @claude review always to subscribe the PR to reviews on each push.
• What it flags. Commit a REVIEW.md and CLAUDE.md to the repository. Anthropic's docs say REVIEW.md is read from the base or default branch, never from the pull request's own changes.

Automatic reviews skip drafts and pull requests from forks, though @claude review still works on them.

What Are the Limits of Claude Code Review?

Claude Code Review comments and reports, and stops there. Its boundaries are worth understanding before you depend on it:

• It does not approve or block pull requests. Anthropic's docs say a review never approves or blocks a PR, and the check run does not conclude as a failure, so it cannot gate a merge on a critical finding.
• It does not push fixes. Inside a review, access to the repository is read-only and a review does not push commits. Resolving each finding is on the developer, or on a separate tool.
• Plan gated. Team and Enterprise only for the managed service, with the data-retention exceptions above.
• Cost scales with volume and depth. No flat per-review price, and "After every push" multiplies reviews per PR. A monthly spend limit caps the total, but once it is reached reviews are skipped until the next period.
• Minutes, not seconds. Anthropic says most reviews finish in about 6 to 10 minutes since GA.
• No model choice. Code Review chooses the models reviews run on; you adjust thoroughness through the review level instead.

None of these make it a bad tool. They define where it fits. Teams that want a reviewer that can also gate merges, auto-approve safe changes and fix bugs will need something beyond comments.

How does Claude Code Review compare with Macroscope?

The practical differences are availability, cost per review, and what the reviewer can do beyond commenting. Claude Code Review comments. Macroscope comments, and can also auto-approve, fail a check, and open fix PRs, on usage-based pricing for any workspace.

Dimension	Claude Code Review	Macroscope
Availability	Team and Enterprise organizations	Any workspace, $100 in usage credit to start
Pricing	Token usage via usage credits; launch average $15 to $25 per review	$0.05 per KB of diff; 76 cents average on Balanced, 55 cents on Budget
Spend controls	Monthly spend limit for the service	Caps per review and per PR on automatic reviews, monthly budget limits, auto-review off by repo or label
Context	Agents read the diff and surrounding code	Review looks past the diff into the rest of the repository for breakage in dependent code
Auto-approve safe PRs	No	Yes, Approvability
Fail a check on critical issues	No, the check run does not conclude as a failure	Yes, Check Run Agents with conclusion: failure
Auto-fix bugs	No	Yes, Fix It For Me opens fix PRs and iterates until CI passes
Custom rules	REVIEW.md and CLAUDE.md	Check Run Agents defined in plain English
Local review before push	/code-review in Claude Code	Macroscope CLI, including a Claude Code plugin (/macroscope:codereview)
Our 118-bug benchmark	Not tested	48.3% detection, 57 of 118 (benchmark)

Both are serious reviewers, and both leave humans in control of merges by default. The distinction is leverage. Macroscope can auto-approve the safe PRs so reviewers skip them, fail a check to block a risky one, and open a fix PR with Fix It For Me, closing the loop instead of adding another comment thread. And it is reachable for individuals and small teams, not gated to Team or Enterprise plans.

How do Claude Code Review, CodeRabbit, Greptile and Macroscope compare?

On pricing model and measured detection, the four most-searched AI code review tools for GitHub differ more than their feature lists suggest. Detection numbers for Macroscope, CodeRabbit and Greptile come from our 118-bug benchmark; Claude Code Review was not part of that benchmark set. Competitor prices are read from each vendor's own pricing page.

	Claude Code Review	Macroscope	CodeRabbit	Greptile
Pricing model	Token usage; launch average $15 to $25 per review	$0.05 per KB of diff; 76 cents average on Balanced, 55 cents on Budget; no seat fee	$24 per developer per month billed annually on Essentials ($30 billed monthly), plus optional continued reviews	$1 per Base review credit, no seat fees
Who can buy	Team and Enterprise organizations	Any workspace	See vendor	See vendor
Bug detection (our benchmark)	Not tested	48.3% (57 of 118)	45.8% (54 of 118)	23.6% (17 of 72 evaluated)
Approves or blocks PRs	No	Yes (Approvability, Check Run Agents)	See vendor	See vendor
Opens fix PRs	No	Yes (Fix It For Me)	See vendor	See vendor

For pricing detail, see CodeRabbit pricing and Greptile pricing. For deeper head-to-heads, see CodeRabbit vs Macroscope, Macroscope vs Greptile, and the best AI code review tools for GitHub in 2026 roundup.

How Do I Run a Claude Code Code Review on Any Plan?

Run /code-review inside a Claude Code session: it reviews your current branch and uncommitted changes for correctness bugs, on any plan, and counts toward your normal plan usage or API billing. It is the Claude Code code review option for developers who are not on Team or Enterprise, or who want feedback before they open a PR.

Per Anthropic's docs, you can point it at something other than your current branch by typing a target: a pull request number such as /code-review 1234, a branch comparison such as /code-review main...my-feature, or a path such as /code-review src/auth. Add --fix to have Claude apply the findings to your working tree, or --comment to post them on a GitHub pull request as inline comments. The review runs as a background subagent and follows your CLAUDE.md, but not REVIEW.md, which only configures the managed service.

The limitation is the same one any self-review has: it is the same family of model, often in the same session, reviewing code it may have just written. That is why many teams pair it with a separate reviewer. The Macroscope CLI installs an optional plugin for Claude Code, so /macroscope:codereview runs a Macroscope review on your branch and Claude Code validates and fixes the confirmed findings. Our guide to AI code review in Claude Code, Codex, Cursor and OpenCode walks through it, the best AI code review CLI tools roundup compares the terminal options, and the Claude Code GitHub guide covers the Claude Code GitHub Action and @claude mentions.

How do you review the pull requests Claude Code generates?

Put a separate reviewer on every agent-written pull request, because the volume of code goes up faster than human review capacity. A growing reason teams want AI code review is the PRs that coding agents write. When Claude Code, Codex or Copilot opens pull requests, the volume of code goes up and the human review burden goes up with it. Agent-written code tends to look clean on the surface, which makes it easy to approve without catching what it missed.

That is a review problem, not a generation problem. The reviewer needs to look beyond the diff to catch the duplicated utility the agent did not know existed, the edge case that is not in the repo's history, and the regression in code that depends on the change. Macroscope's review looks past the diff into the rest of the repository for exactly that, and the metric that matters shifts from how much code the agent produced to how much of it actually shipped. See Landed vs. All and custom AI code review agents for how to enforce your own standards on every agent PR. If you run several Claude Code sessions at once, see how to run multiple Claude Code agents in parallel for keeping each one's diff separate and reviewable.

What are the best Claude Code Review alternatives?

The main alternatives are Macroscope, CodeRabbit and Greptile, plus /code-review in Claude Code itself for review before you push.

If Claude Code Review is not a fit, because you are not on Team or Enterprise, because $15 to $25 per review is hard to budget at your PR volume, or because you want a reviewer that can approve, block and fix, these are the alternatives worth evaluating:

• Macroscope. Usage-based AI code review at $0.05 per KB of diff (averaging 55 cents on Budget and 76 cents on Balanced), available to any workspace, with auto-approve (Approvability), merge-blocking custom checks, and auto-fix (Fix It For Me). Detected 48.3% of bugs (57 of 118) on our benchmark. Suits teams that want review and action without per-seat fees.
• CodeRabbit. Per-developer AI code review, $24 per developer per month on Essentials billed annually. See CodeRabbit pricing and CodeRabbit alternatives.
• Greptile. $1 per Base review credit, no seat fees. See Greptile alternatives.

How do you get started with Claude Code Review or an alternative?

If you are on Team or Enterprise, turn on Claude Code Review for a few pilot repositories; otherwise start with /code-review locally and a usage-based reviewer on GitHub.

Claude Code Review is a strong option if you are already on a Claude Team or Enterprise plan and comfortable budgeting token-based reviews that averaged $15 to $25 at launch. If you are not, or if you want a reviewer that auto-approves the safe PRs, blocks the dangerous ones and fixes bugs instead of only flagging them, Macroscope works for any workspace, prices by the work it does (averaging 55 cents on Budget and 76 cents on Balanced), and starts with $100 in usage credit. Set up GitHub code review in five minutes and run it on your real pull requests.

Frequently Asked Questions

What is Claude Code Review?

Claude Code Review is Anthropic's managed, multi-agent pull request review service for GitHub. It launched as a research preview on March 9, 2026 and became generally available on October 15, 2026 for Team and Enterprise plans. Several agents examine each change and the surrounding code in parallel, a verification step filters out false positives, and Claude posts inline comments ranked by severity, a summary, and a check run named Claude Code Review. It never approves or blocks the pull request, so a human makes the merge decision.

How much does Claude Code Review cost?

There is no flat per-review price. Claude Code Review is billed on token usage through usage credits, separately from your plan's included usage, and each review's cost scales with the review level, the size of the change and the complexity of the surrounding code. Anthropic's launch announcement said reviews generally average $15 to $25. An organization can cap the total with a monthly spend limit.

Does Claude Code Review have a spend limit?

Yes. An organization can set a monthly spend limit for the Claude Code Review service in its usage settings. When that limit, the organization's overall spend limit, or its usage credit balance runs out, reviews are skipped and the pull request gets a comment explaining why. Reviews resume at the next billing period or sooner if the limit is raised.

Can I use Claude Code code review on a Pro or Max plan?

Not the managed service, which is for Team and Enterprise organizations. But the /code-review command works on any plan: run it in a Claude Code session to review your branch, a pull request number, a branch comparison or a path, add --fix to apply findings, or add --comment to post them on the pull request. It counts toward your normal plan usage or API billing.

Can Claude Code Review approve or block pull requests?

No. Anthropic's documentation says a review never approves or blocks a pull request, and its check run does not conclude as a failure. If you want a reviewer that can auto-approve safe PRs and fail a check to block merges on critical issues, Macroscope offers both through Approvability and Check Run Agents with conclusion: failure.

Does Claude Code Review fix bugs automatically?

No. Inside a review, Claude Code Review has read-only access to the repository and does not push commits, so resolving each finding is up to the developer. For automatic fixes, Macroscope's Fix It For Me creates a branch, commits a fix, opens a PR and iterates until CI passes.

How do I trigger Claude Code Review on a pull request?

Each repository has a setting for when reviews run: once when a pull request opens, after every push, or manually. In any mode, someone with write access can comment @claude review to run a single review, or @claude review always to subscribe that pull request to a review on each later push. Automatic reviews skip drafts and pull requests from forks.

How long does a Claude Code Review take?

Anthropic's GA announcement says most reviews finish in about 6 to 10 minutes, roughly 3x faster than the 20 minutes typical during the research preview. Larger and more complex changes, and the Deep review level, take longer.

What is the difference between Claude Code and Claude Code Review?

Claude Code is Anthropic's coding agent, which you run in a terminal or other client to write and edit code, and which includes the /code-review command for reviewing your own branch. Claude Code Review is the managed service that reviews GitHub pull requests automatically on Anthropic's infrastructure for Team and Enterprise organizations. Claude Code is something you run; Claude Code Review is something that runs on every PR for your team.

How do I set up Claude Code Review on GitHub?

An Owner of your Claude organization turns on "Allow code review" in Admin settings under Claude Code on claude.ai, connects GitHub, installs the Claude GitHub App, and chooses repositories, then clicks Finish. There are no workflow files to add. Then set a review level (Quick, Standard or Deep) and when reviews run for each repository, and verify with a test pull request. For automated GitHub code review on any workspace, Macroscope is a five-minute setup with $100 in usage credit.

What are the best Claude Code Review alternatives?

The most common alternatives are Macroscope, CodeRabbit and Greptile. Macroscope is usage-based (averaging 55 cents on Budget and 76 cents on Balanced), available to any workspace, and can auto-approve, block and fix; it detected 48.3% of bugs (57 of 118) on our benchmark. CodeRabbit is priced per developer, from $24 per developer per month billed annually, and Greptile charges $1 per Base review credit with no seat fees. For review before you push, /code-review in Claude Code works on any plan.

Claude Code Review vs Macroscope: which should I choose?

Choose Claude Code Review if you are already on Claude Team or Enterprise and want comment-only, multi-agent review billed on token usage. Choose Macroscope if you want review that also acts, auto-approving safe PRs, blocking risky ones and fixing bugs, on usage-based pricing (averaging 55 cents on Budget and 76 cents on Balanced) for any workspace. The deciding factors are plan availability, cost per review, and whether you need the reviewer to do more than comment.