Claude Code Review exits preview and runs about 3x faster

Anthropic on October 15 took Code Review out of research preview and made it generally available on Claude Team and Enterprise plans, with most reviews now finishing in 6 to 10 minutes, down from about 20. Eligible Enterprise organizations can also claim at least $10,000 in Code Review credits by January 15, 2027, and since the credits expire 30 days after the claim, claiming them before any repository is running reviews burns part of that window.
An admin switches it on in Claude Code settings, installs the GitHub App and selects the repositories, then sets a review level for each one: Quick, Standard or Deep, so routine codebases get a quick pass and the most critical ones a deep review. Developers set nothing up; reviews run automatically on every new pull request. The credit terms are on a separate support page.
The speed gain is measured against the preview: about 3x faster for most reviews. Each finding now states the problem, what triggers it and the fix in a few sentences. Both changes answer what preview teams told Anthropic: Code Review was catching real issues before a human reviewer looked at the PR, but they wanted reviews that were quieter, sized to each change and faster. Anthropic also runs it on its own pull requests, which it says shows where reviews slow its engineers down.
On bug-finding, Anthropic says Code Review now catches more of the bugs that matter but gave no figures for how many more, so the reference point is still its preview launch post: on Anthropic’s own pull requests, the share that got substantive review comments rose from 16% before Code Review to 54% with it, and engineers marked fewer than 1% of findings as incorrect.
Code Review is billed on token usage. At the preview launch Anthropic said reviews generally average $15 to $25, scaling with PR size and complexity, and called it a more thorough and more expensive option than its Claude Code GitHub Action, which stayed open source and available. Admins could also set a monthly spending cap for the organization from the start. At that average, the minimum $10,000 credit covers roughly 400 to 670 reviews, and they would all have to run within the 30 days after the claim. Some commenters on the launch questioned whether a 20-minute review at that price would be practical for high-volume engineering workflows, InfoQ reported, and some said the pricing may limit adoption for smaller teams.
Each review is run by a team of agents that search for bugs in parallel, verify them to filter out false positives and rank them by severity; the result lands on the PR as one overview comment plus inline comments on specific bugs. Cat Wu, Anthropic’s head of product, said the tool goes after logic errors rather than style and gives a light security analysis, with Claude Code Security doing the deeper one, per TechCrunch. Code Review does not approve pull requests: that, Anthropic says, is “still a human call”.
Up next
Claude now picks its own agent team in Managed Agents
You’re stuck in GitHub approving agent PRs all day, and Anthropic’s answer is flows, where Claude plans the work and writes the program that runs its subagents. Right now, who splits your agents’ work, you or the model?
Related stories
- Claude Code team hands most code reading to Claude
- Sonnet 5.5 cut CodeRabbit’s review bill by about 60%
- Claude Code 2.1.285 lets admins lock down API providers
- Claude Code 2.1.281 adds Bedrock guardrails to its gateway
- Claude gets write access to Microsoft 365 on Dec 9
- Claude Code stops wiping .env files
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

