CodeCrew
A coordinated team of AI agents that plan, build, regression-check, and QA software — and refuse to sign off without physical proof.
Planner → Builder → Checker → QA
Specialized agents per role, each with the narrowest tool surface they need. No single-agent monoliths.
Physical proof, not vibes
Changes ship only after real curl output, real test runs, and real query results are collected.
Three-fix protocol
For risky changes, CodeCrew generates three distinct candidates and scores each before touching code.
Built on Claude
Powered by Claude Opus, Sonnet, and Haiku — tiered by task, priced by importance.