jenasuraj/Ai_agents ? reverse-engineered prompt

Build me this AI agents hub in Python, with a clean shared core and separate experimental agents that can be run individually. I want a supervisor style agent that can take a user request, figure out which worker agents are needed, route the task through them in order, call tools when needed, and then combine everything into one final answer.

Include a few useful agents like web research and scraping, podcast style content with text to speech, stock and news analysis, GitHub automation, Notion help, travel planning, retrieval based Q and A, and a more advanced orchestration flow. Keep the structure modular so new agents can be added without rewriting everything, and make the shared request and response format consistent across the project.

If it helps, look up current docs online for LangGraph, LangChain, and MCP so the wiring is modern and correct. I also want a simple way to run at least the supervisor flow from the command line, and a path toward a FastAPI backend or small dashboard later.