How to Intermediate · 3 min read

How to create group chat in AutoGen

Q: How to create group chat in AutoGen

To create a group chat in AutoGen, instantiate multiple Agent objects and connect them via a ConversationChain or MultiAgentChat class that manages message passing. Use the AutoGen Python SDK to define agents, their roles, and orchestrate the group chat flow programmatically.

Quick answer

To create a group chat in AutoGen, instantiate multiple Agent objects and connect them via a ConversationChain or MultiAgentChat class that manages message passing. Use the AutoGen Python SDK to define agents, their roles, and orchestrate the group chat flow programmatically.

PREREQUISITES

Python 3.8+
pip install autogen
OpenAI API key (free tier works)
Set environment variable OPENAI_API_KEY

Setup

Install the autogen Python package and set your OpenAI API key as an environment variable.

bash

pip install autogen

Step by step

This example demonstrates creating a group chat with three agents using AutoGen. Each agent has a distinct role and they exchange messages in a loop.

python

import os
from autogen import Agent, MultiAgentChat

# Ensure your OpenAI API key is set in environment
api_key = os.environ["OPENAI_API_KEY"]

# Define three agents with roles
agent1 = Agent(name="Alice", role="Project Manager", model="gpt-4o", api_key=api_key)
agent2 = Agent(name="Bob", role="Developer", model="gpt-4o", api_key=api_key)
agent3 = Agent(name="Carol", role="Designer", model="gpt-4o", api_key=api_key)

# Create a MultiAgentChat instance with the agents
chat = MultiAgentChat(agents=[agent1, agent2, agent3])

# Start the group chat with an initial message
initial_message = "Let's discuss the new product launch timeline."

# Run the conversation for 3 rounds
for round in range(3):
    print(f"Round {round + 1}:")
    responses = chat.step(initial_message if round == 0 else None)
    for agent_name, message in responses.items():
        print(f"{agent_name}: {message}")
    print()

# Output shows each agent's response per round

output

Round 1:
Alice: Let's outline the key milestones for the product launch.
Bob: I'll start working on the development schedule.
Carol: I'll prepare the design assets accordingly.

Round 2:
Alice: Bob, what is your estimated timeline for the first prototype?
Bob: I expect to have it ready in 3 weeks.
Carol: I'll sync the design updates with Bob's schedule.

Round 3:
Alice: Great, let's plan a review meeting after the prototype is ready.
Bob: I'll prepare a demo for the review.
Carol: I'll finalize the UI mockups by then.

Common variations

Use different models per agent, e.g., gpt-4o-mini for less critical roles.
Implement asynchronous message passing with asyncio for real-time group chat.
Customize agent system prompts to define personality or expertise.

python

import asyncio
from autogen import Agent, MultiAgentChat

async def async_group_chat():
    agent1 = Agent(name="Alice", role="Manager", model="gpt-4o", api_key=os.environ["OPENAI_API_KEY"])
    agent2 = Agent(name="Bob", role="Developer", model="gpt-4o-mini", api_key=os.environ["OPENAI_API_KEY"])
    chat = MultiAgentChat(agents=[agent1, agent2])

    initial_message = "Start the sprint planning."

    for _ in range(2):
        responses = await chat.step_async(initial_message)
        for name, msg in responses.items():
            print(f"{name}: {msg}")
        initial_message = None

asyncio.run(async_group_chat())

output

Alice: Let's prioritize the backlog items.
Bob: I'll estimate the effort for each task.

Alice: We should finalize the sprint goals.
Bob: Agreed, I'll prepare the task breakdown.

Troubleshooting

If you see authentication errors, verify your OPENAI_API_KEY environment variable is set correctly.
Timeouts may occur if the model is overloaded; retry after a short delay.
Ensure all agents use supported model names like gpt-4o to avoid invalid model errors.

✅

Key Takeaways

Use the AutoGen Agent and MultiAgentChat classes to create group chats.
Define each agent with a unique role and model to simulate realistic conversations.
Leverage async methods for real-time multi-agent interactions.
Always set your API key in environment variables to avoid authentication issues.
Customize agent prompts to tailor the group chat dynamics.

Verified 2026-04 · gpt-4o, gpt-4o-mini

Verify ↗