Anthropic launches Claude Code tool to review AI-generated code

Anthropic is adding a new layer to Claude Code as AI-generated pull requests start piling up faster than teams can properly review them.

The company on Monday announced Code Review for Claude Code, a research preview for Team and Enterprise customers. In Anthropic’s description, opening a pull request sends multiple agents across the code in parallel to look for bugs, check likely findings, and sort issues by severity. The review then shows up on the PR as one summary comment with inline notes where something looks wrong. 

Anthropic says this is the same kind of system it already uses on nearly every pull request internally, and it is built for a deeper read rather than a quick skim.

Anthropic says the pressure behind this is straightforward. Its engineers are producing a lot more code, and review is becoming the part that slows everything down. The company says code output per engineer has grown 200% over the past year, while the share of pull requests receiving substantive review comments rose from 16% to 54% after it started using this system internally.

Also read: Anthropic says it will no longer automatically pause development when a model could be dangerous.

TechCrunch says Anthropic product head Cat Wu said enterprise customers kept running into the same problem as Claude Code produced more pull requests, and that “Code Review is our answer to that.” Wu also said Anthropic chose to “focus purely on logic errors” because developers get frustrated when AI feedback is “not immediately actionable.”

That is what makes this launch more interesting than another coding feature announcement. Anthropic is not just saying Claude can help write code. It is saying the next problem comes after that, when more code shows up than humans can realistically read with care. 

Code Review is Anthropic’s attempt to move into that part of the workflow, from generation into verification. Anthropic says reviews usually take around 20 minutes and cost about $15 to $25 depending on the size and complexity of the pull request.