The surge of AI-generated code is creating a new problem: a bottleneck in review. As developers use tools like Anthropic's Claude Code to produce software at unprecedented speeds, the traditional process of human peer review is struggling to keep up. On Monday, Anthropic introduced a direct response: an automated AI reviewer designed to scan this flood of code for critical errors before it ships.

Cat Wu, Anthropic's head of product, explained the catalyst. "Enterprise leaders keep asking us, 'Claude Code is creating a mountain of pull requests. How do we review them efficiently?'" The new product, Code Review, integrates directly with GitHub, automatically analyzing submissions and leaving detailed comments on potential logic errors and fixes.

The system, now in research preview for Teams and Enterprise customers, employs a multi-agent architecture. Several AI agents examine code simultaneously from different angles, with a final agent synthesizing the findings. It prioritizes logic flaws over style, labeling severity with a color code—red for critical, yellow for review-worthy, purple for historical issues. "We focus purely on logic errors to catch the highest-priority things," Wu said, noting that vague feedback frustrates developers.

This launch arrives as Anthropic leans into its enterprise business, where subscriptions have quadrupled this year. Claude Code's run-rate revenue has exceeded $2.5 billion. The tool is a premium offering; Wu estimates each review costs $15 to $25 on average, a price justified by the scale of the need. "This is coming from an insane amount of market pull," she said. The goal is to let companies like Uber and Salesforce, already using Claude Code, build faster and with fewer bugs than ever before.

Source: TechCrunch