Open Code Review (OCR for short) is an open-source quality gate tool for AI-generated code, positioned as "the first open-source CI/CD quality gate built specifically for AI-generated code." It detects defects in code generated by AI coding assistants (such as Copilot, Cursor, and Claude) that traditional tools tend to miss, including hallucinated imports, phantom dependencies, outdated APIs, over-engineering, and security anti-patterns. The tool is powered by a local LLM or any OpenAI-compatible model provider, is free and self-hostable, and supports 6 languages.
Key Capabilities OCR's core purpose is detecting defects unique to AI-generated code. The material notes that code generated by AI coding assistants contains defects that traditional tools are completely unable to find, and OCR can detect these issues across 6 languages. It includes built-in language-specific detectors for 6 languages, along with a hallucinated package database (covering npm, PyPI, Maven, and Go modules). Its deep scan (L3) combines three layers of analysis for maximum coverage. In addition, OCR provides an AI auto-fix feature, ocr heal, which lets AI automatically fix discovered issues and allows changes to be reviewed before being applied. The tool also supports generating a complete interactive HTML report.
How to Use OCR offers three usage forms: MCP Server, CLI, and CI/CD Action.
CLI: It can be run with npx @opencodereview/cli scan src/, requiring zero configuration.
MCP Server: It can be integrated directly into AI IDEs via the Model Context Protocol, supporting Claude Desktop (configured in claude_desktop_config.json), Cursor, Windsurf, and VS Code Copilot (by adding the same configuration in MCP settings). Available MCP tools include: ocr_scan (quality gate scan), ocr_heal (AI auto-fix), and ocr_explain (issue explanation). The material also mentions that the OCR MCP Server follows the standard Model Context Protocol and can be paired with Google's Chrome DevTools MCP Server to form a complete AI-native development workflow—oneinspects the running application, while the other inspects the source code.
CI/CD Integration: GitHub Actions integration is provided, and the material says configuration can be completed in 30 seconds.
Model Support: Ollama can be used for 100% local analysis, or any remote LLM provider can be connected, with a consistent interface.
Use Cases The material lists the target users: teams using AI coding assistants (Copilot, Cursor, Claude Code, Codex, or any LLM tool that generates production code); open-source maintainers who need to review hallucinated imports, outdated APIs, and security anti-patterns in AI-generated PRs before merging; DevOps/platform engineers who want to add a quality gate to their CI/CD pipeline without sending code to cloud services; security-conscious teams that want to run fully locally (Ollama) and keep code on their own machines; and independent developers seeking a free, fast, zero-configuration solution.
Limitations and License OCR uses the BSL-1.1 license, is free for personal and non-commercial use, and will convert to Apache 2.0 on March 11, 2030. Commercial use requires a Team or Enterprise license. The material does not provide specific pricing, billing details, a complete list of language-specific detectors, the specific technical details of the three-layer analysis, performance data, or complete comparisons with other tools.
