Code Review for Claude Code

4 hours ago

Claude Code introduces a new Code Review feature, modeled after Anthropic's internal system, now in research preview for Team and Enterprise.
The Code Review system uses a team of agents to deeply analyze PRs, catching bugs that human reviewers might miss, increasing substantive review comments from 16% to 54%.
Agents work in parallel to identify, verify, and rank bugs by severity, providing a high-signal overview and in-line comments on PRs.
Review depth scales with PR size and complexity, with average reviews taking about 20 minutes.
Internal testing shows 84% of large PRs (over 1,000 lines) receive findings, averaging 7.5 issues, while small PRs (under 50 lines) see 31% with 0.5 issues on average.
A case study highlights a critical bug caught in a one-line change that would have broken authentication, demonstrating the system's value.
Early access customers, like TrueNAS, have benefited from identifying latent bugs in adjacent code during refactors.
Code Review is more thorough and expensive than lighter solutions, averaging $15–25 per review, with controls for admins to manage spend and usage.
Available now in beta for Team and Enterprise plans, with detailed documentation available.

Hasty Briefsbeta