CodeDecay

CodeDecay is a local-first CLI and GitHub Action that audits pull requests created by AI coding agents. It identifies weak tests, regression risks, and maintainability issues by mapping changed files to affected APIs, routes, authentication, and data schemas. The tool runs deterministic analysis without requiring cloud services or LLM calls, and generates evidence bundles for agents like Claude Code, Codex, and Cursor to fix issues in closed-loop workflows.

Designed for developers using AI coding assistants, CodeDecay separates runtime proof from static test references and mocked boundaries. It supports explicit verification through configured tests, builds, and product checks, with results surfaced as JSON and SARIF for local editors. The makers report it caught 23 of 23 planted issues in benchmarks with no false positives and zero cost.