Claude Code Review: what it costs and how to turn it on

Claude Code Review is Anthropic's automated pull request reviewer, and since October 15 it has been generally available to Claude Team and Enterprise organizations. There is no flat price. Each review is billed on token usage through usage credits, and during the research preview Anthropic said reviews averaged $15 to $25 each.

What Anthropic announced on X

@ClaudeDevs posted the news Thursday evening: "Code Review is now generally available for Claude Team and Enterprise plans." The post says reviews are faster and catch more of the bugs that matter, and that admins can now set a Quick, Standard or Deep level per repository. It had passed a million views by Sunday, so plenty of engineering teams were paying attention.

Anthropic's GA announcement backs up the speed claim with numbers. Most reviews now finish in 6 to 10 minutes, compared with about 20 during the preview. Eligible Enterprise organizations can claim at least $10,000 in Code Review credits by January 15, 2027, and the credits expire 30 days after you claim them.

How a review works

When a pull request opens on GitHub, several agents read the diff and the code around it in parallel, looking for logic errors, security holes and broken edge cases. A verification step checks each candidate finding against the code and throws out false positives. The findings that survive are deduplicated, ranked by severity and posted as inline comments, with a summary and a check run called Claude Code Review.

If you're nervous about giving an AI the merge button, two details should help. A review can't approve or block a pull request, so people still decide what ships. And its access to the repository is read-only: it reads the code without building or running it, and it doesn't push commits. Out of the box it hunts for bugs that would break production and ignores formatting nits.

What Claude Code Review costs

The costs page in the docs says plainly that there is no per-review price. Your bill comes down to two things:

• How much each review costs. That scales with the review level, the size of the change, the complexity of the surrounding code and how many candidate issues need checking.
• How many reviews run. Each repository can review once when a PR opens, after every push, or only when someone comments @claude review.

Watch that second setting. New repositories default to "After every push," so every push on every PR is another billed review. Reviews are charged to the organization, outside your plan's included usage, and they land on your Anthropic bill even if your developers reach Claude through Bedrock or another cloud.

For scale, Anthropic's March launch post is where the $15 to $25 average comes from. That post also said 16% of Anthropic's own PRs got substantive review comments before the tool and 54% did after, with engineers marking less than 1% of findings as incorrect. Those are Anthropic's numbers about Anthropic's code, so read them as a vendor claim. The GA post gives no new average. My guess is that the Quick level comes in under it, but nobody has published that figure.

How to turn it on

An Owner of the Claude organization turns on Code Review in the Claude Code admin settings, installs the Claude GitHub App, picks repositories and sets a default review level. Developers don't have to do anything after that, because reviews appear on new PRs automatically. The docs recommend a pilot before a wide rollout: set a monthly spend limit, enable a few repositories in "once" or manual mode, then take the average cost per review from the admin table and multiply it by your monthly PR count.

Some organizations can't use it. Code Review works only with GitHub, and it isn't available to organizations with Zero Data Retention, the HIPAA configuration or customer-managed encryption keys. If you aren't on Team or Enterprise, you can still run the /code-review command inside a Claude Code session on any plan.

My read

I care less about the speedup than about how Anthropic packaged this. Review is its own metered step, with its own budget and its own verifier, and it has no power to merge. As AI writes more of the code, checking that code becomes the slow part. A reviewer that costs a known amount and leaves the final call to a human is a much easier sell to a cautious manager than yet another autonomous agent.

What this means for a small business

Most Fox Valley businesses don't have a code review process. They have a contractor, a website and some scripts nobody has opened in two years. The idea still applies: as AI takes on more of the work, put a cheap, predictable check between what it produces and your customer, and keep a person on the final decision. If you'd like help figuring out where that check belongs in your operations, our free process audit covers exactly that.