Microsoft Agent Framework Documentation
Microsoft Agent Framework is a comprehensive SDK for building AI agents in .NET and Python. It provides abstractions for AI integration (Azure AI, OpenAI, Anthropic, Ollama), tool execution, Model Context Protocol (MCP), durable agents with state persistence, workflows, and multi-channel adapters (Teams, WebChat, DirectLine).
Quick Reference
| Topic | Entry Point |
|-------|-------------|
| Project overview | references/README.md |
| FAQ & nightly builds | references/docs/FAQS.md |
| .NET samples | references/dotnet/samples/README.md |
| .NET getting started | references/dotnet/samples/GettingStarted/README.md |
| Python samples | references/python/samples/README.md |
| Architecture decisions | references/docs/decisions/README.md |
| Durable agents TTL | references/docs/features/durable-agents/durable-agents-ttl.md |
| Azure Functions hosting | references/dotnet/samples/AzureFunctions/README.md |
| Workflows | references/dotnet/samples/GettingStarted/Workflows/README.md |
When to use
Use this skill when the user asks about:
- Building AI agents with Microsoft Agent Framework
- .NET or Python agent development
- MCP (Model Context Protocol) servers and clients
- Durable agents and session state management
- Agent tools and tool execution
- Teams, WebChat, or DirectLine adapters
- Agent-to-agent (A2A) communication
- Azure AI, OpenAI, Anthropic, or Ollama integration
- AG-UI (Agent UI) protocol implementation
- Workflows and orchestration patterns
- Azure Functions agent hosting
How to find information
- First, read
references/STRUCTURE.mdto see all available documentation files - Identify the relevant section/files based on the user's question
- Read specific files for detailed information
STRUCTURE.md contains a complete file listing organized by directory - always check it first before searching.
Key Directories
dotnet/samples/GettingStarted/- Step-by-step .NET tutorialsdotnet/samples/AzureFunctions/- Serverless hosting examplesdotnet/samples/GettingStarted/AgentProviders/- LLM provider integrationdotnet/samples/GettingStarted/Workflows/- Workflow patternspython/samples/getting_started/- Python tutorialsdocs/decisions/- Architecture Decision Records (ADRs)