Anthropic Launches AI Code Review Tool

Peer feedback is essential in coding for early bug detection, maintaining codebase consistency, and enhancing overall software quality.

The emergence of "vibe coding"—using AI tools that generate code from plain-language instructions—has transformed developer workflows. While accelerating development, these tools have also introduced new bugs, security vulnerabilities, and poorly understood code.

Anthropic's solution is an AI reviewer designed to catch bugs before they enter the codebase. The new product, named Code Review, launched Monday in Claude Code.

"We've seen significant growth in Claude Code, particularly within the enterprise sector. A recurring question from enterprise leaders is: Now that Claude Code is generating numerous pull requests, how do we ensure they are reviewed efficiently?" said Cat Wu, Anthropic's Head of Product, to TechCrunch.

Pull requests are a standard mechanism for developers to submit code changes for review before integration. Wu noted that Claude Code has dramatically increased code output, leading to a surge in pull request reviews and creating a bottleneck in the deployment pipeline.

"Code Review is our answer to that challenge," Wu stated.

Anthropic's launch of Code Review—initially available in research preview for Claude for Teams and Claude for Enterprise customers—comes at a critical time for the company.

On Monday, Anthropic filed two lawsuits against the Department of Defense following the agency's designation of Anthropic as a supply chain risk. This dispute may lead Anthropic to rely more heavily on its rapidly growing enterprise business, where subscriptions have quadrupled since the start of the year. According to the company, Claude Code's run-rate revenue has exceeded $2.5 billion since its launch.

"This product is specifically targeted at our larger enterprise users—companies like Uber, Salesforce, and Accenture that already use Claude Code and now need help managing the volume of pull requests it generates," Wu explained.

She added that development leads can enable Code Review by default for their entire team. Once activated, it integrates with GitHub, automatically analyzing pull requests and posting direct comments on the code to highlight potential issues and suggest fixes.

The focus is on identifying logical errors rather than stylistic concerns, Wu emphasized.

"This is crucial because developers often find AI-generated feedback frustrating when it's not immediately actionable. We decided to concentrate solely on logic errors to address the highest-priority fixes," she said.

The AI explains its reasoning step-by-step, detailing the identified issue, why it could be problematic, and potential solutions. The system categorizes issue severity using a color code: red for critical, yellow for issues requiring review, and purple for problems related to pre-existing code or historical bugs.

Wu explained that the process is fast and efficient, utilizing multiple specialized agents that examine the code from different perspectives. A final agent consolidates and ranks the findings, removes duplicates, and prioritizes the most important items.

The tool includes a basic security analysis, and engineering leads can customize additional checks based on internal best practices. Wu noted that Anthropic's more recent product, Claude Code Security, offers a more in-depth security analysis.

The multi-agent architecture makes this a resource-intensive product, Wu acknowledged. Similar to other AI services, pricing is token-based and varies with code complexity—though Wu estimated the average cost per review would be between $15 and $25. She described it as a premium solution necessary as AI tools produce increasing amounts of code.

"Code Review is a direct response to massive market demand," Wu said. "As engineers use Claude Code to reduce friction in feature development, the need for code review has intensified. We are confident this will enable enterprises to build faster than ever before, with significantly fewer bugs."