sentrux

Why we included this project

The usual way to judge an AI coding agent is by its tests and whatever a human catches in review, which makes it easy to miss slow declines in code quality. Sentrux takes a different approach: it scans a repository in 52 languages, compresses five graph-theoretic metrics like modularity and acyclicity into a single quality score, and serves that score through an MCP server so an agent can read it mid-session and know what to refactor next. The gate command saves a baseline before an agent runs, so teams using Claude Code, Cursor, or any MCP client can fail CI if the session drags the score down. Because the whole thing is a single pure-Rust binary with no runtime dependencies, setup is minimal and it drops into existing workflows. That makes it a practical feedback mechanism for anyone who wants agent-written code to measurably improve rather than quietly decay.