AMA - Find, route, and secure every AI agent on your machine

I ran a scanner on my dev machine and found 457 AI agent assets across Claude Code, Codex, and Pi Agent — most completely unmanaged. API keys in plaintext, no spend tracking, permissions everywhere. AMA started as a personal tool to clean up my own mess. It grew into four things: scan (discover every agent), route (right model for each task), track (spend per agent), and audit (security report). Open source because agent sprawl is about to be everyone's problem, and we need shared infrastructure to solve it. Would love honest feedback — especially on what's missing or broken.