autogen

What it is and what it does

Autogen is a programming framework for building AI agents that work together to solve tasks. It provides a set of agent types (ConversableAgent being the core) that can send and receive messages, generate replies using language models or tools, and be orchestrated into multi-agent systems using patterns like group chats, swarms, and nested chats. The framework handles agent-to-agent communication, tool registration and execution, human-in-the-loop workflows, and integration with various LLMs.

The package is built on pydantic for configuration, httpx for HTTP requests, and tiktoken for token counting. It depends on python-dotenv for API key management and termcolor for terminal output. This is the original AutoGen repository, now in maintenance mode receiving security and critical bug fixes only; the actively developed version is published separately as ag2.

Use it for:

• Build a multi-agent code review system where one agent writes code and another reviews it without rewriting.
• Create a lesson planning workflow with specialized agents for planning, curriculum review, and feedback.
• Implement autonomous workflows where agents collaborate to research, analyze, and summarize information.
• Develop human-in-the-loop systems where AI agents handle tasks but escalate decisions to human reviewers.
• Orchestrate complex problem-solving by having agents with different expertise work together on a shared goal.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

A framework for building AI agents that communicate with each other and with language models to solve tasks through multi-agent collaboration and orchestration patterns.

Yes, if you need a multi-agent framework and accept maintenance-mode status. The package is production-stable, actively maintained for security, has low install friction, and supports Python 3.10+. However, be aware it receives only critical fixes and security patches—no new features. If you need active development and new capabilities, consider ag2 instead.

Install

autogen on PyPI

pip install autogen

uv add autogen

poetry add autogen

Installing autogen

Before you install

Low friction installation with a pure Python wheel. Actively maintained with recent releases; marked as production-stable but now in maintenance mode (security and critical fixes only), with the actively developed version available separately as ag2.

License in practice

Licensed under Apache-2.0 (permissive), allowing commercial and private use with minimal restrictions; suitable for most commercial projects.

Quickstart

pip install autogen[openai]

from autogen import ConversableAgent, LLMConfig

llm_config = LLMConfig.from_json(path="OAI_CONFIG_LIST")
agent = ConversableAgent("agent", llm_config=llm_config)
response = agent.run(recipient=other_agent, message="task")

Requires Python >= 3.10 and an LLM API configuration (e.g., OpenAI key in OAI_CONFIG_LIST file).

Verify before relying

• Whether the maintenance-mode status affects long-term stability for new projects or only blocks new features.
• Performance characteristics when orchestrating many agents or handling large conversation histories.
• Whether tiktoken dependency is required for all use cases or only for token counting with certain LLMs.

Package facts

Evidence: autogen-0.14.1-py3-none-any.whl

Keywords: ag2, ag2.ai, ag2ai, agent, agentic, ai, autogen

Tags

More Software Development packages

Further reading