openqodex/openqodex

What it solves

OpenQodex provides an automated, multi-stage code review process that runs locally before you push code. It bridges the gap between static analysis tools and AI coding agents, ensuring that code changes are vetted by both deterministic scanners and an AI reviewer to catch bugs, security vulnerabilities, and linting issues.

How it works

OpenQodex operates in two primary stages. First, it runs specialized scanners (like Semgrep or Gitleaks) on your changed files to identify potential issues. Second, it tasks an AI reviewer—such as Claude Code or Codex—to evaluate these findings. The AI must provide a reason for either accepting or rejecting each scanner finding. The system integrates with your local development environment or coding agent, acting as a "push gate" that can block commits if they fail to meet your configured severity thresholds.

Who it’s for

It is designed for developers who use AI coding agents (like Claude Code, Cursor, or Cline) and want to add a rigorous, automated quality-control layer to their workflow without relying on external, opaque cloud-based review services.

Highlights

• Automated "push gate" that prevents pushing code with unreviewed or high-severity issues.
• Multi-stage validation combining traditional static analysis with LLM-based reasoning.
• Supports 22 built-in scanners for secrets, security, dependencies, and linting.
• Extensible architecture allows users to add custom scanners from GitHub with a secure trust-based approval process.
• Generates comprehensive reports in multiple formats, including HTML, Markdown, JSON, and SARIF.