VRSEN/agency-swarm ? reverse-engineered prompt

Build me a Python framework for running a small agency of AI agents that can work together like a real team. I want to be able to create agents with clear roles like CEO, assistant, or developer, give each one its own instructions, files, and tools, then control which agents are allowed to talk to which other agents. It should feel reliable and ready for real projects, not just a toy demo.

Please make it easy to add custom tools with validation, and if possible let me generate tools from an OpenAPI schema too. I also want optional thread persistence so conversations can be saved and loaded across sessions. Give me a simple terminal experience and a basic browser demo so I can try it right away.

Include a clean starter example, solid docs, and tests. Assume I will use an OpenAI API key from an env file. If anything is unclear, look up the current OpenAI Agents SDK docs and make sensible choices.