Anthropic has introduced a new code review tool for its Claude AI system, allowing developers to analyze code directly within the platform without relying on additional plugins. The feature is currently available as a research preview.
The tool can be enabled through Claude’s admin settings under the “Code Review” option. Once activated, it deploys a “team of agents” that scans a codebase for potential bugs.
Automated Bug Detection
According to Anthropic, the system searches for issues in parallel, verifies each potential bug to eliminate false positives, and ranks the findings by severity.
The company said the tool is modeled on the code review system used internally at Anthropic. It argues that manual code reviews often become a bottleneck for engineers working on software projects.
Claude’s automated review process does not approve pull requests on its own. Instead, the company said the system is designed to assist developers by narrowing down issues so reviewers can focus on the code that is being shipped.
Mixed Reactions From Developers
Early feedback from developers has been mixed. The head of AI at Hedgineer, a hedge fund asset management software company, said on X that the feature offers little improvement over an existing setup using Claude’s GitHub integration.
They said the main difference is that the new tool does not require a trigger, such as a GitHub Action, and can be configured directly within the Claude desktop environment.
Another developer, Claude coder Thariq, said the tool uses significantly more compute resources and tends to detect more complex bugs.
Higher Cost
Anthropic acknowledged that the tool may be more expensive than some alternatives. Code reviews are billed based on token usage and typically cost between $15 and $25, with larger pull requests costing more.
The company said the system prioritizes deeper analysis, which can increase compute usage compared with other solutions, including its open source GitHub Action.
