Anthropic Details AI-Powered Code Review for Claude Code

Low-risk changes require no human approval, while high-risk ones need dual review. Review cycles are capped at three rounds to prevent excessive feedback. The process improves efficiency: one-review-pass rate rose from 65% to 78%, and five-or-more-rounds cases dropped from 8.9% to 1.5%. Human reviewers remain responsible for architecture, product direction, and final outcomes.

Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.