AI-Powered Bug Detection in Pull Requests: Introducing New Claude Tool

Post Views: 4

Claude Code Review Launched by Anthropic

A new AI-powered tool, Claude Code Review, has been launched by Anthropic to help identify bugs in pull requests. This tool utilizes a team of artificial intelligence agents to examine every pull request, providing a thorough review that can catch issues that human reviewers might miss.

How Claude Code Review Works

The system operates by dispatching multiple agents to search for bugs in parallel. The findings are then verified to filter out false positives, and confirmed issues are ranked according to their severity. The agents do not approve pull requests, but instead provide a comprehensive overview of the findings directly on the pull request, with specific bugs noted in inline comments.

Effectiveness of Claude Code Review

According to Anthropic, the effectiveness of Claude Code Review has been demonstrated through internal testing. In large pull requests involving over 1,000 lines of changed code, the tool identified issues in 84% of cases, with an average of 7.5 issues per review. In smaller pull requests involving fewer than 50 lines of code, the tool identified issues in 31% of cases, with an average of 0.5 issues per review. Moreover, the company reported that engineers agreed with the findings in almost all cases, with less than 1% of findings being marked as incorrect.

Availability and Pricing

Claude Code Review is available as a research preview beta for Team and Enterprise plans, and is billed based on token usage, scaling with the size and complexity of the pull request, as well as the number of issues requiring verification. The average cost per review ranges from $15 to $25. Administrators can set a total monthly spending limit across reviews, enable reviews only for selected repositories, and track reviewed pull requests, acceptance rates, and total costs.

Enabling Claude Code Review

To enable the feature, administrators need to install the GitHub app in Claude Code settings and select the repositories to be reviewed. Once enabled, reviews are run automatically on new pull requests without additional configuration.

Comparison to Existing Tools

Claude Code Review is a more comprehensive option than the existing Claude Code GitHub Action, which remains open source and available. The new tool provides a deeper analysis of pull requests, making it a valuable resource for teams seeking to improve the quality and security of their code.