Reference · Glossary

AutoGen

Last updated

A Microsoft framework for multi-agent apps built around **conversable agents** — agents that talk to each other (and optionally a human) in a chat loop until the task is done, with built-in support for code execution and human-in-the-loop checkpoints.

#When to use

Workflows where agents genuinely need to converse and critique each other's output (e.g. a coder agent and a reviewer agent going back and forth), or where you want a human approval step wired into the loop.

#When not to

A fixed, linear pipeline with a known order of steps — a simpler sequential chain (or LangGraph) is easier to reason about than an open-ended conversation loop.

#Example

from autogen import ConversableAgent
coder = ConversableAgent("coder", llm_config={"model": "gpt-4o-mini"})
reviewer = ConversableAgent("reviewer", llm_config={"model": "gpt-4o-mini"})
coder.initiate_chat(reviewer, message="Write a function to dedupe a list, then have it reviewed.")