↓ Skip to main content
Side Effects
  1. posts/

Alibaba's Open Code Review mixes fixed rules with an LLM agent

··View as Markdown·
Link Code-Review Ai Claude Code

Source: alibaba/open-code-review on GitHub

Alibaba’s Open Code Review doesn’t trust the LLM alone. A fixed pipeline checks for null pointers, thread safety, XSS and SQL injection, and the agent writes line comments on top of that. It runs as a plugin in Claude Code, Codex, Cursor and Kimi Code, and in Delegation Mode it reuses the host agent’s model instead of asking for its own API key. The README calls it “battle-tested at Alibaba’s scale” and doesn’t show a single example review.

TL;DR

  • Go, Apache-2.0, ~43,000 GitHub stars
  • Hybrid design: deterministic pipelines + LLM agent, comments on specific lines
  • Built-in rules: null pointers, thread safety, XSS, SQL injection
  • LLM: OpenAI and Anthropic compatible APIs
  • Plugins: Claude Code, Codex, Cursor, Kimi Code, QCA Forward
  • Needs Git 2.41+ and ocr CLI: npm install -g @alibaba-group/open-code-review
  • First run: ocr config provider, ocr config model, ocr llm test
  • Delegation Mode: host agent’s model reviews, no setup (QCA Forward uses it)
  • Claude Code: /plugin marketplace add alibaba/open-code-review, then /plugin install open-code-review@open-code-review
  • Commands: /open-code-review:review, /open-code-review:delegate-review

Related