I tested whether locking interface contracts helps AI app rebuilds (github.com)

A new tool has been developed that focuses on generating a reliable rebuild specification (CLAUDE.md) from existing applications, allowing coding agents to rebuild without ambiguity. This MCP server analyzes a codebase, extracting routes, configurations, and tests, and systematically enforces discipline through interface contracts. Initial tests validate its performance against a real code repository, successfully managing two independent agent transitions across different model tiers. This development is significant for the AI/ML community as it addresses critical pain points in the application rebuild process, including behavioral equivalence issues highlighted in prior research. By locking interface contracts before running tests and implementing a disciplined, one-at-a-time testing approach, the tool aims to substantially improve rebuild accuracy and reliability. Key technical features include an ambiguity resolution queue, automated signal tracking, and a mutation testing process that ensures generated tests effectively catch introduced faults. This methodological advancement could enhance the efficiency and dependability of AI-driven coding processes, minimizing the risks of silent bugs and improving developer workflows.