Our take
- Activity
- 91
- Community
- 83
- Issues
- 80
- Pull requests
- 47
+3,433 stars in the last 10 days
- Releases more often than 93% of the projects we track
- More contributors than 77% of the projects we track
Overview
Open Code Review is a command-line tool for reviewing Git changes with an LLM agent. Deterministic steps handle file selection and rule matching, while the agent can inspect full files and search repository context to produce structured, line-level findings.
It is aimed at developers and teams who want repeatable AI-assisted reviews rather than relying on a general-purpose coding agent to decide what to inspect. It also supports whole-file scans and a delegation mode in which a coding agent performs the review.
Key Features
- Reviews workspace changes, branch ranges, and individual commits.
- Scans complete files or directories, including when there is no useful Git diff.
- Uses deterministic file selection, file bundling, and rule matching alongside an LLM agent.
- Gives the agent tools to read files and search repository context.
- Produces structured review comments with line-level positions.
- Supports configurable model providers, plus delegation to a coding agent without OCR-managed LLM configuration.
- Provides integrations for coding agents and CI/CD systems.
Use Cases
- Development teams can add AI-assisted checks to code review workflows and CI pipelines.
- Individual developers can review staged, unstaged, or untracked changes before submitting them.
- Maintainers can scan unfamiliar repositories or directories for issues without relying on a diff.
- Users of supported coding agents can delegate review work to their existing agent setup.
Getting Started
Install the CLI with npm:
npm install -g @alibaba-group/open-code-review
Then configure a model provider and run ocr review in a Git repository. See the README and official documentation for setup options and usage details.
Alternatives
- ai-outfitter/actions: Runs configurable AI code-review profiles as GitHub Actions, while open-code-review combines deterministic review steps with an LLM agent.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| open-code-review | Go | Apache-2.0 | 43.5k | Not checked yet |
| actions | Shell | MIT | 2 | New |
Considerations
- Git 2.41 or newer is required.
- OCR-managed reviews require configuring an LLM provider and model. Delegation mode is available when a supported coding agent should perform the review instead.
- The project describes a precision-oriented trade-off: its benchmark summary says recall is lower than that of general-purpose agents, so it may miss some issues in favor of fewer false alarms.
- Review quality and API usage depend on the configured model and review workflow.
Found this useful?
Share it with someone who would like open-code-review.