Microsoft AutoGen was created by Microsoft Research for multi-agent conversations and has evolved significantly since its inception.
| Year | Milestone |
|---|---|
| 2023 | Initial release, rapid community adoption |
| 2024 | 40,000+ GitHub stars, AutoGen Studio launched |
| 2025 | v0.7.5 release, Microsoft Agent Framework announced |
| 2026 | 55.3k+ GitHub stars, transition to Agent Framework |
AutoGen started as a research project at Microsoft Research focused on conversational multi-agent systems. The key innovation was enabling agents to communicate through natural language conversations to solve complex tasks.
The new Microsoft Agent Framework builds on AutoGen’s learnings:
| Feature | AutoGen | Agent Framework |
|---|---|---|
| Graph-based Workflows | ⚠️ Limited | ✅ Full support |
| Data Flows | ⚠️ Conversational | ✅ Explicit data passing |
| Streaming | ✅ | ✅ Enhanced |
| Checkpointing | ⚠️ Manual | ✅ Built-in |
| Human-in-the-Loop | ✅ | ✅ Enhanced |
| Time-Travel Debugging | ❌ | ✅ |
| OpenTelemetry | ⚠️ Manual | ✅ Built-in |
| Multi-Language | Python, .NET | Python, .NET (unified) |
| Component | Purpose |
|---|---|
| AutoGen | Main framework |
| AutoGen AgentChat | High-level agent API |
| AutoGen Studio | Visual builder |
| AutoGen Bench | Benchmarking |
| AutoGen.Net | .NET implementation |
| Extension | Purpose |
|---|---|
| autogen-ext[openai] | OpenAI integration |
| autogen-ext[azure] | Azure OpenAI integration |
| autogen-ext[anthropic] | Anthropic integration |
| autogen-ext[ollama] | Ollama (local LLM) |
| autogen-ext[mcp] | Model Context Protocol |
AutoGen is built on these principles:
Microsoft provides official migration guides: