Dev.to
7/8/2026
The original title is "Wardrail: an independent referee for AI-generated code in your CI pipeline"
Original: AI writes my code now. Here's how I stopped rubber-stamping it.
Short summary
A solo founder launches Wardrail, a tool that independently reviews AI-generated code by enforcing project-specific contracts on every PR. It combines deterministic AST-level security scans with an independent LLM verdict to produce a trust score and precise proof of compliance, running entirely in your own CI with zero-knowledge key handling. The tool also builds a native code graph and ships an MCP server so agents like Claude Code and Cursor can query project structure and contract rules while writing code.
- •Wardrail enforces project contracts on AI-generated PRs via AST security scans plus independent LLM verdicts
- •Runs in your own CI with zero-knowledge key handling and read-only GitHub scope
- •Includes a code graph, MCP server for Claude Code/Cursor/Windsurf, and 238 reusable agent skills and loops
Generated with AI, which can make mistakes.
Is this a good recommendation for you?
