Anthropic's Claude Code Review Tool Automatically Detects Pull Request Vulnerabilities

Anthropic has introduced a new AI-powered code review tool, Code Review, on its Claude Code platform. It is designed to automatically flag potential issues before code is merged, helping to ease efficiency bottlenecks in enterprise development caused by soaring review demands. Launched on Monday, the feature is currently in a research preview for Claude for Teams and Claude for Enterprise customers.

The rise of "vibe coding," where developers use AI to generate code through natural language, has significantly boosted productivity. However, it also introduces challenges like code vulnerabilities, security risks, and reduced readability. In an interview with TechCrunch, Anthropic product lead Cat Wu noted that while Claude Code accelerates code generation, it has led to a rapid increase in Pull Requests. Traditional manual reviews struggle to keep up, creating a major bottleneck in software delivery.

The new Code Review feature integrates with GitHub to automatically analyze submitted Pull Requests, directly annotating code with potential issues and suggested fixes. It focuses on identifying logical errors rather than stylistic concerns to ensure feedback is actionable. The AI explains its reasoning step-by-step, detailing the problem's location, potential impact, and possible solutions. It uses a color-coded system to indicate risk: red for critical issues, yellow for items needing attention, and purple for problems related to historical errors or existing code structures.

Technically, the system employs a multi-agent collaboration mechanism. Several agents review the codebase in parallel from different angles, and a summarizer agent consolidates the findings, removes duplicates, and prioritizes issues for a more efficient automated process. The tool also includes basic security analysis, allowing companies to add custom rules based on internal standards. For deeper security checks, Anthropic offers its previously released Claude Code Security tool.

Given the computational demands of the multi-agent architecture, Code Review uses a token-based pricing model, with costs varying by code complexity. Cat Wu estimates the average review cost falls between $15 and $25. She emphasized that the product targets large enterprise clients like Uber, Salesforce, and Accenture, which already use Claude Code extensively and need automated review to manage high volumes of Pull Requests.

This launch comes at a pivotal time for Anthropic. On the same day, the company filed two lawsuits against the U.S. Department of Defense regarding its designation as a supply chain risk. Concurrently, its enterprise business is expanding rapidly; the company reports that enterprise subscription numbers have quadrupled since the start of the year, and Claude Code's annualized revenue has surpassed $2.5 billion since its debut.

As AI-generated code becomes more prevalent, automated code review is emerging as essential infrastructure for enterprise development. With this feature, Anthropic aims to help companies reduce vulnerability risks while accelerating development, pushing AI-assisted software development toward greater reliability.