AI Output Verifier for Codebases
VeriPatch is an open-source desktop application that helps developers verify AI-generated code changes against a real local repository before those changes are trusted, committed, or merged.
Instead of generating more code, VeriPatch answers a more important question:
Can this AI-generated change be trusted in this specific codebase?
- Open a local repository and paste a unified diff, import a
.patchfile, inspect uncommitted changes, or review same-repo pull requests - Run a local verification pipeline (compile check, lint, tests, security scan, dependency audit)
- Optionally use an AI model (via OpenRouter) for risk analysis and review
- Get a clear Safe / Risky / Broken verdict with explanations
- Merge or close an open pull request after review from the desktop app
- Export markdown reports for pull requests or code review
Early development — not yet ready for production use.
- Rust (1.85+)
- Node.js (20+)
- macOS, Linux, or Windows
- Optional: GitHub CLI (
gh) for pull request review, merge, and close actions
npm install --prefix crates/veripatch-app/frontend
cargo build --releasecargo run -p veripatch-appcargo run -p veripatch-cli -- --helpcrates/
veripatch-app/ # Tauri desktop application with a React frontend
veripatch-cli/ # Command-line interface
veripatch-core/ # Core verification engine
veripatch-ai/ # AI provider integration (OpenRouter)
veripatch-runners/ # Check runners (compile, lint, test, etc.)
veripatch-rules/ # Security & pattern detection rules
veripatch-report/ # Report generation (markdown, JSON)
docs/ # Documentation
examples/ # Example diffs and usage
See CONTRIBUTING.md for guidelines.
Licensed under either of:
at your option.
See SECURITY.md for reporting vulnerabilities.