Abilityai/trinity ? reverse-engineered prompt

Build me a self hosted AI agents platform I can run on my own server with Docker.

I want a web app where I can create agents, give them names, connect them to Claude Code, Codex, or Gemini, and run them in isolated containers. It should let me start and stop agents, schedule jobs, watch what each agent is doing in real time, and see a clear audit trail of everything that happened. I also want basic team features, like multiple users, shared sessions, handoffs between agents, and a simple approval flow before risky actions happen.

Please make the setup as easy as possible with a quick start that works locally, then a production ready path using Postgres and Docker Compose. If anything important is unclear, look up the current docs online and follow the repo’s existing patterns. Focus on a polished dashboard, sensible defaults, and something that feels ready to use instead of a demo.