Microsoft Agent Framework Documentation
by Microsoft
The official docs for Microsoft's successor to AutoGen and Semantic Kernel — agents, harnesses, and graph-based multi-agent workflows in .NET, Python, and Go.
Overview
Microsoft Agent Framework is the direct successor to both Semantic Kernel and AutoGen, built by the same teams, and its Microsoft Learn documentation is the canonical reference for it. The docs organize the framework into three capability categories: Agents (individual LLM-driven agents that call tools and MCP servers, with provider support for Microsoft Foundry, Anthropic, Azure OpenAI, OpenAI, Ollama and more), Harness (an opinionated batteries-included agent for long multi-step tasks, with planning and todo tracking, context compaction, file access and memory, don't-ask-again tool approval, and observability), and Workflows (graph-based orchestration connecting agents and functions with type-safe routing, checkpointing, and human-in-the-loop support). Underneath sit the foundational building blocks the docs cover in depth — model clients for chat completions and responses, an agent session for state management, context providers for agent memory, middleware for intercepting agent actions, and MCP clients for tool integration. Language pivots let you read every page in C#, Python, or Go (Go is in public preview), and dedicated migration guides walk existing Semantic Kernel and AutoGen users across. The framework combines AutoGen's simple single- and multi-agent abstractions with Semantic Kernel's enterprise features — session-based state management, type safety, filters, telemetry, and broad model and embedding support. The docs are actively maintained (the overview page was last updated in July 2026) and include a notable piece of engineering advice up front: if you can write a function to handle the task, do that instead of using an AI agent.
At a Glance
- Topic
- Frameworks
- Level
- Intermediate
- Format
- Documentation
- Cost
- Free
- Duration
- Self-paced reference; ~2-3 hours for the get-started tutorial
- Provider
- Microsoft
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Build your first agent with the agent-framework package in Python, .NET, or Go, then add tools, multi-turn conversations, and memory
- ✓Choose correctly between an autonomous agent and an explicit graph workflow, using the framework's own decision table
- ✓Wire agents to tools and MCP servers, including hosted MCP tools, via the built-in MCP clients
- ✓Orchestrate multi-agent systems as graph-based workflows with type-safe routing, checkpointing, and human-in-the-loop steps
- ✓Use the Harness agent's planning, todo tracking, context compaction, and tool-approval features for long-running tasks
- ✓Apply middleware, agent sessions, and context providers for state management, telemetry, and observability in production
- ✓Migrate an existing Semantic Kernel or AutoGen codebase onto Agent Framework using the official migration guides
Highlights
- •The official successor to AutoGen and Semantic Kernel — the forward path for anyone on either framework
- •Same docs, three languages: every page pivots between C#, Python, and Go
- •Covers the production concerns most agent tutorials skip: checkpointing, middleware, telemetry, human-in-the-loop
- •Actively maintained by Microsoft with a 180-day update cycle
Who It’s For
Best For
- ✓AI engineers building multi-agent systems on .NET or Python in an enterprise environment
- ✓Teams currently on AutoGen or Semantic Kernel planning a migration
- ✓Azure and Microsoft Foundry developers who need agents with type safety and telemetry
- ✓Engineers evaluating agent frameworks against LangGraph, CrewAI, or the OpenAI Agents SDK
Prerequisites
- •Working knowledge of Python, C#/.NET, or Go
- •Familiarity with LLM APIs and tool/function calling
- •Basic understanding of agent loops and MCP is helpful
FAQ
What is Microsoft Agent Framework Documentation?
Official Microsoft Learn documentation for Agent Framework (MAF), an open, multi-language framework for building production AI agents and multi-agent workflows. Aimed at .NET, Python, and Go engineers who need enterprise-grade agent orchestration with state management, telemetry, and human-in-the-loop control.
Is Microsoft Agent Framework Documentation free?
Microsoft Agent Framework Documentation is free to access.
What level is Microsoft Agent Framework Documentation for?
Microsoft Agent Framework Documentation is aimed at a intermediate audience. Recommended background: Working knowledge of Python, C#/.NET, or Go, Familiarity with LLM APIs and tool/function calling, Basic understanding of agent loops and MCP is helpful.
How long does Microsoft Agent Framework Documentation take?
Expect roughly Self-paced reference; ~2-3 hours for the get-started tutorial. Most learners work through it at their own pace.
What will I learn from Microsoft Agent Framework Documentation?
You'll learn: Build your first agent with the agent-framework package in Python, .NET, or Go, then add tools, multi-turn conversations, and memory; Choose correctly between an autonomous agent and an explicit graph workflow, using the framework's own decision table; Wire agents to tools and MCP servers, including hosted MCP tools, via the built-in MCP clients; Orchestrate multi-agent systems as graph-based workflows with type-safe routing, checkpointing, and human-in-the-loop steps; Use the Harness agent's planning, todo tracking, context compaction, and tool-approval features for long-running tasks; Apply middleware, agent sessions, and context providers for state management, telemetry, and observability in production; Migrate an existing Semantic Kernel or AutoGen codebase onto Agent Framework using the official migration guides.