What it does
Open Code Review is a command-line tool that began as Alibaba Group's internal AI code review assistant and is now open source under Apache-2.0. Its ocr command reads Git diffs, sends the changed files to a configurable LLM through an agent with tool access, and returns structured review comments pinned to specific lines. The agent can read whole files, search the codebase and look at other changed files for context. The design splits the work between fixed engineering steps and the agent. File selection, grouping of related files into isolated sub-agent review units, rule matching and comment positioning are handled deterministically, while the agent decides what extra context to fetch. The README reports higher precision and F1 than a general-purpose agent using the same model on its own benchmark, with lower recall. You supply the model endpoint, either a built-in provider or a custom OpenAI- or Anthropic-compatible one. Reviews can cover the working tree, a branch range or a single commit, and ocr scan reviews whole files. JSON output, CI integrations for GitHub Actions, GitLab CI, GitFlic CI and Gerrit, and a browser session viewer are included. A delegation mode lets your own coding agent perform the review, with plugins listed for Claude Code, Codex, Cursor, Kimi Code and OpenCode.
Capabilities
- Line-level review comments from Git diffs, commits or branch ranges
- ocr scan for reviewing whole files when there is no useful diff
- Custom review rules with path filtering
- Delegation mode that lets your coding agent do the review
- JSON output and CI integrations for GitHub Actions, GitLab CI, GitFlic CI and Gerrit
- Browser session viewer to replay reviews and mark comments as fixed or ignored
When to use it
- Adding automated review comments to pull requests in CI
- Reviewing your own local changes before opening a pull request
- Auditing an unfamiliar codebase or directory that has no meaningful diff
- Teams that want review rules tailored to their own code
