Open Code Review

Open Code Review is an open-source artificial intelligence agent designed to automate software code reviews. Originally developed and tested on over one million real-world tasks by Alibaba Group, the tool operates as a local command-line interface (CLI) application. It reads repository files, searches codebase context, and analyzes code diffs to detect bugs, security risks, and logic errors. Developers can connect the agent to any major Large Language Model (LLM) API, including those from OpenAI, Anthropic, DeepSeek, and private self-hosted models. This bring-your-own-LLM approach ensures that proprietary source code remains private and conforms to internal security policies.

The system utilizes a hybrid architecture that separates deterministic engineering from LLM reasoning. Deterministic modules manage file filtering, task routing, and exact line-number positioning, while the LLM focuses exclusively on semantic analysis and risk detection. This structure minimizes hallucinated feedback and significantly reduces token consumption compared to standard LLM queries. The agent outputs precise, line-level comments directly to the terminal or continuous integration pipeline. It supports more than 40 programming languages and allows users to adjust the review depth to balance processing speed with analysis thoroughness.