microsoft/autogen ? reverse-engineered prompt

I want to build a framework in Python that allows me to create and manage multiple AI agents that can work together. The main idea is that I can define different agents with specialized roles, like a "researcher" that can browse the web or a "writer" that is good at summarizing text.

I should be able to create a primary "manager" agent that can receive a complex task and then delegate sub-tasks to the specialized agents. They need to be able to communicate with each other to solve the problem. Hook this up to use OpenAI models for their intelligence.

Also, build a simple web interface or "studio" on top of this framework. This UI should let me visually define the agents, their skills, and their workflows without having to code everything from scratch, making it easy to prototype different multi-agent setups.