Anthropic is making Code Review available for Claude Code, an AI system that automatically analyzes code and detects bugs before someone has to do it manually.
Development teams are increasingly using AI coding tools, leading to a sharp rise in the number of pull requests. Reviewing code is therefore taking up more and more time and causing burnout for some developers. With Code Review, Anthropic aims to solve this problem by having AI automatically analyze code before it reaches a developer.
Multiple AI agents check code
The new feature uses a system where different AI agents simultaneously analyze a codebase. Each agent examines the code from a different perspective, looking for potential errors or vulnerabilities.
An additional agent then collects all findings and ranks the issues based on their severity. The system marks critical errors in red, potential issues in yellow, and unresolved issues in purple. For every problem found, the AI describes what could go wrong and suggests a possible solution.
Reviews take longer and cost tokens
The extensive analysis does mean that reviews take a relatively long time. According to Anthropic, it takes an average of about twenty minutes to analyze a pull request. Furthermore, costs are calculated based on token usage. A code review costs between 15 and 25 dollars on average, depending on the size and complexity of the pull request.
More bugs found in large code changes
Anthropic says the system has been used internally for several months. In large pull requests of more than a thousand changed lines of code, the AI found issues in 84 percent of reviews, averaging about 7.5 per review. Code Review is currently available in research preview for Claude Code Team and Enterprise subscribers.