Alibaba Open-Sources 'Open Code Review', an AI Code Review Tool Internally Validated for Two Years

Alibaba has open-sourced its internal AI code review tool under the name Open Code Review, after a process involving tens of thousands of developers and the identification of millions of code defects. This tool adopts a hybrid architecture combining Deterministic Engineering and LLM Agents, addressing issues such as positional errors and insufficient coverage inherent in pure language-based agents.

Core Architecture and Performance

Existing general-purpose AI agents have limitations in reviewing large-scale changes, such as reviewing only some files or misaligning line references. Open Code Review ensures essential accuracy in steps like file selection, bundling, and rule matching through engineering logic, delegating only dynamic decision-making to the LLM. According to AACR-Bench benchmark results, it achieves significantly higher Precision and F1 scores compared to the same model, reduces token consumption to approximately 1/9 of the level, and improves review speed. However, due to its design characteristics, Recall is relatively low.

Key Features and Usage

This tool supports Windows, macOS, and Linux, and integrates with major AI coding agents such as Claude Code, Codex, and Cursor. It can be installed via npm and operates in environments with Git 2.41 or higher.

• Review Mode: Performs reviews across various scopes, including workspace, branch range, and single commits.
• Full Scan: The ocr scan command allows auditing the entire repository without Git history.
• Delegation Mode: OCR performs only file selection and rule interpretation, while the actual review is handled by the LLM of the connected AI coding agent. In this case, no separate LLM configuration is required.
• Built-in Rules: Provides default multilingual security and quality rules, including NPE, thread safety, XSS, and SQL injection.

License and Compatibility

It is distributed under the Apache-2.0 license and is compatible with OpenAI and Anthropic APIs. It has recorded 22.3k stars on GitHub and is under active development.