Capability · Framework — agents

AutoGen

AutoGen (now at v0.4+) is a major rewrite of Microsoft's pioneering multi-agent framework. It offers a layered architecture — a Core event-driven runtime, an AgentChat API for conversational agents, and Extensions for tools and models. AutoGen Studio provides a drag-and-drop UI, while Magentic-One ships as an opinionated multi-agent system for browsing, coding, and file tasks.

Framework facts

Category: agents
Language: Python / .NET
License: CC-BY-4.0 / MIT
Repository: https://github.com/microsoft/autogen

Install

pip install autogen-agentchat autogen-ext[openai]

Quickstart

import asyncio
from autogen_agentchat.agents import AssistantAgent
from autogen_ext.models.openai import OpenAIChatCompletionClient

async def main():
    model = OpenAIChatCompletionClient(model='gpt-4o')
    agent = AssistantAgent('assistant', model_client=model)
    print(await agent.run(task='Say hi in Hindi.'))

asyncio.run(main())

Alternatives

CrewAI — role-based alternative
LangGraph — lower-level graph-based agents
OpenAI Agents SDK — single-vendor OpenAI alternative
Semantic Kernel — Microsoft's sibling .NET-first SDK

Frequently asked questions

Is AutoGen still actively developed?

Yes — Microsoft Research shipped a full rewrite as v0.4 with async and event-driven primitives, and the team continues to release features like Magentic-One and AutoGen Studio. A separate community fork also exists under the ag2ai name.

AutoGen or Semantic Kernel?

AutoGen is Python/.NET and multi-agent first. Semantic Kernel is a general-purpose enterprise SDK with planning and plugins. Microsoft is progressively aligning the two — pick AutoGen for research-style multi-agent work, SK for enterprise app integration.

Sources

AutoGen — docs — accessed 2026-04-20
AutoGen — GitHub — accessed 2026-04-20