agentscope-ai/AgentTeams ? reverse-engineered prompt

Build me an open source app for coordinating multiple AI agents in a shared chat room, where a person can see everything happening and step in whenever needed.

I want a manager agent that assigns work to several worker agents, keeps track of the conversation, and makes the whole process auditable and transparent. The experience should feel like a real team working in a Matrix style room, with a normal chat client for the human side and clear visibility into who said what and why. It should also support sharing files between agents, and make it easy to mix different kinds of workers, like ones that plan, ones that write code, and ones that do autonomous tasks.

Make it practical to run in an enterprise setting, with secure central control, easy onboarding, and a clean way to define teams and workers with config. If you need to check current docs or best practices for Matrix or Kubernetes style deployment, look them up online.