Environment setup

Welcome! I am Param, and in this workshop we are going to build a travel support assistant with AutoGen GroupChat. By the end, you will have specialist agents collaborating under a supervisor, with streaming tool calls surfaced to the frontend.

The assistant can look up bookings, search hotels, check flight status, book and cancel reservations, answer policy questions, and even convert currency through an external MCP server. You will start with one agent and one tool, then grow the system into a team.

What you will build

A streaming FastAPI service backed by AutoGen agents, tools, and a supervisor router.

terminal
bash
# Clone the workshop repository
git clone https://github.com/learnwithparam/multi-agent-autogen.git
cd multi-agent-autogen

# One command to set up everything
make dev

This clones the repo, creates a virtual environment with uv, installs AutoGen and FastAPI, and starts the server on http://localhost:8000.

.env
bash
# Copy the example and pick one provider
cp .env.example .env

# OpenRouter (has a free tier)
LLM_PROVIDER=openrouter
OPENROUTER_API_KEY=your_key_here
OPENROUTER_MODEL=openai/gpt-4o-mini

# Or OpenAI
# LLM_PROVIDER=openai
# OPENAI_API_KEY=sk-...
# OPENAI_MODEL=gpt-4o-mini

AutoGen speaks the OpenAI API, so any OpenAI-compatible endpoint works. Pick whichever provider you already have a key for. Tool calling must be supported by the chosen model.

Validation checklist: Environment setup checklist

Loading practice…