Anthropic launches AI-powered Code Review tool for Claude Code at $15-25 per review

Mar 9, 2026

Anthropic has launched Code Review, an AI-powered tool for its Claude Code platform, to address the bottleneck of reviewing AI-generated code. Using a multi-agent system, it analyzes pull requests for logical bugs, costing $15-25 per review. The tool, aimed at enterprise users, has been met with some developer criticism over its high price but showed strong results in internal testing, tripling the rate of substantive feedback on code submissions.

Claude Code Review launch

▪The Code Review tool is designed to find bugs in AI-generated code by automatically analyzing pull requests on GitHub.
▪Code Review is available in a research preview for customers on Claude Team and Enterprise plans.
▪Anthropic launched its "Code Review" tool, a new feature for the Claude Code platform, on March 9, 2026.
▪The launch comes as Anthropic faces a DoD supply chain risk designation, increasing its reliance on its enterprise business.

Pull request workflow context

▪The trend of "vibe coding," using plain language AI prompts to generate code, has increased development speed but also introduced new bugs and risks.
▪The rise of AI coding assistants has created a "pull request bottleneck" as human reviewers are overwhelmed by the volume of code submissions.
▪A pull request is a process where developers submit code changes for review before they are merged into a project's main codebase.

Multi-agent review system architecture

▪A final agent aggregates the findings from parallel agents, removes duplicates, and ranks the issues by importance.
▪An average review takes about 20 minutes, with the time scaling based on the pull request's complexity.
▪Identified issues are color-coded by severity: red for critical, yellow for potential problems, and purple for issues related to pre-existing code.
▪Code Review uses a multi-agent system where several AI agents analyze code in parallel from different perspectives.
▪The system is designed to focus on logical errors rather than stylistic issues to provide more actionable feedback for developers.

Bug detection test results

▪For large pull requests over 1,000 lines, the tool found issues 84% of the time, averaging 7.5 issues.
▪In Anthropic's internal tests, using Code Review increased the rate of pull requests with substantive comments from 16% to 54%.
▪Anthropic engineers marked fewer than 1% of the tool's findings as incorrect during internal usage.

Per-review pricing structure

▪Some developers have criticized the price as expensive, with one noting a single review can cost as much as a monthly subscription to other tools.
▪The tool is billed based on token usage, with an average review costing between $15 and $25.
▪A less expensive, open-source Claude Code GitHub Action tool for code review also remains available from Anthropic.
▪The cost scales with the size and complexity of the code being reviewed.