Anthropic: Code Review, an Expensive and Exclusive Solution

⚡

Key Takeaways

1Anthropic has launched Code Review to manage the influx of AI-generated code in open source projects.

2The system improves the pull request review rate from 16% to 54%, but at a high cost.

3The pricing of $15 to $25 per use limits access to large teams, excluding small projects.

💡Why it matters — Small open source maintainers are left behind, exacerbating inequalities in access to advanced tools.

Anthropic and Its New Tool Code Review

Anthropic has recently introduced Code Review, a multi-agent artificial intelligence system designed to address the proliferation of AI-generated code contributions in open source projects. This new tool has been created to analyze and evaluate code contributions, a need that has become crucial with the massive increase in these contributions.

A Leap in Code Review Efficiency

By utilizing multiple specialized agents for each task, Code Review has managed to increase the review rate of pull requests (PR) from 16% to 54%. This system allows for better management of PRs, which are often left pending due to a lack of sufficient human resources.

A Cost That Excludes Small Projects

Although Code Review is intended to alleviate the burden on development teams that are often overwhelmed, its use is charged at between $15 and $25 per review. This pricing primarily targets professional teams, complicating access for small maintainers of open source projects who do not have the same financial means.

An Open Source Community Under Pressure

On March 9, 2026, Anthropic announced the launch of Code Review, a tool that arrives at a crucial time for an open source community overwhelmed by AI-generated code contributions. This announcement comes as the code production per engineer within Anthropic's teams has surged by 200% in a year, thanks to tools like Claude Code.

A Growing Need for Code Review

This impressive increase in productivity has generated a heightened need for code review, while the available time to conduct these reviews is decreasing. Teams find themselves under pressure to maintain code quality while managing a growing volume of contributions.

How Code Review Works

Code Review deploys a team of AI agents to analyze each PR submitted on GitHub. Each agent specializes in a specific task, such as bug detection, false positive verification, and classifying issues by severity. The results include a summary comment and annotations on problematic lines.

Significant but Costly Results

Before the introduction of Code Review, only 16% of PRs received comments. This figure climbed to 54% after its implementation. For significant PRs exceeding 1,000 lines, 84% result in reports, with an average of 7.5 issues identified per review.

The Limitations of Code Review

Despite its advantages, Code Review has notable limitations. Each review takes an average of twenty minutes, which can lead to significant costs for users. Anthropic acknowledges that code review has become a bottleneck for its clients, who complain about the workload overload on developers.

Restricted Accessibility

Code Review is reserved for Team and Enterprise subscriptions of Claude, effectively excluding small open source projects. Although Anthropic offers a free and open source GitHub Action, it does not provide the same level of analysis as Code Review, leaving volunteer maintainers in a difficult position.