The problem: AI coding tools are great but there's no quality control. You get code that "looks right" but fails requirements, has bugs, or doesn't match what you actually asked for.
C-CMCP solves this with a 4-stage pipeline: 1. Claude.ai creates detailed task specs (MUST/SHOULD/COULD requirements) 2. You approve the task 3. Cursor AI implements the code 4. API Claude validates against the original requirements (~$0.01-0.03 per validation) 5. You review validation + test manually 6. Accept or request fixes
We built this in 2.5 hours and it's production-ready. Tested on Windows with PowerShell scripts that automate the workflow.
GitHub: https://github.com/stanpressman/C-CMCP
Built this for my own startup (password manager) but figured others might find it useful. MIT licensed, completely free.
Would love feedback on the approach and what features would make this more useful.