Anthropic has unveiled a "code review" feature that automates review of AI-generated code, TechCrunch reported on March 9.

The code review feature is integrated into the Claude Code platform and helps prevent errors and security vulnerabilities in large codebases in advance.

The code review feature targets enterprise customers, with large companies such as Uber, Salesforce and Accenture as key targets, TechCrunch said.

As developers use AI to speed up code generation, code review is emerging as a new bottleneck. Anthropic is taking this into account by automating code reviews to help companies deploy software.

The system works with GitHub to automatically analyse pull requests and mark logical errors and security issues in the code using colour codes. Red indicates critical errors, yellow indicates issues requiring review and purple indicates conflicts with existing code. It also offers specific fixes so developers can take action immediately.

Anthropic said code review prioritises checking for logical errors and emphasises functional stability over style. AI agents operate in parallel to analyse a codebase from multiple angles, while a final agent consolidates results and sets priorities. This includes security checks. Companies can apply additional validation rules in line with internal policies.

Because the code review feature consumes significant resources, it is also priced high. Anthropic said it expects the average review cost to be $15 to $25, and that it may vary depending on code complexity.