alibaba/open-code-review is a command-line code review tool that reads code changes and produces line-by-line review comments. The latest release is v1.13.3, but the provided record contains no information about what changed in this version. It combines a deterministic static analysis pipeline with an LLM agent to produce review results. The README also describes a feature that inspects entire files, not just the diff.
The tool focuses on finding code defects and attaching review comments to the relevant lines. Its built-in rule set targets multiple languages and includes null pointer exception (NPE), thread safety, XSS, and SQL injection checks. It supports the OpenAI and Anthropic APIs. It is written in Go and released under the Apache-2.0 license.
The README's own benchmark is described as using 50 repositories, 200 real pull requests, 10 programming languages, and 1,505 verified defects, with more than 80 reviewers cross-checking the results. It also reports higher precision and F1, and lower reproducibility, when compared with a general-purpose agent using the same model. The specific scores and comparison conditions are not in the provided record, and these results are the comparison reported in the README. Therefore the difference between precision and reproducibility has to be taken into account, and when judging a performance advantage it is better not to interpret it beyond the scope of this self-comparison.