alibaba/open-code-review

It solves the problem of noisy, incomplete, and imprecise code reviews from general-purpose LLM agents by using a hybrid architecture that ensures full file coverage, accurate line-level comments, and low token consumption. It also addresses the need for a battle-tested, open-source tool that can be integrated into CI pipelines without sacrificing review quality.

Developers can build automated code review pipelines that analyze Git diffs or entire files, catching defects like NPE, thread-safety issues, XSS, and SQL injection. Real-world use cases include CI integration for pull request reviews, auditing unfamiliar codebases, and enforcing coding standards across large teams. The ceiling includes reviewing any codebase with configurable LLM backends (OpenAI, Anthropic) and achieving high precision with low false-positive rates.