
Microsoft Agent Framework is an open-source SDK for developers building AI agents and multi-agent workflows. It connects language models to tools and gives teams control over how agents and functions work together. Python and .NET are supported, with a separate Go SDK in public preview.
Created by the teams behind AutoGen and Semantic Kernel, it combines AutoGen's agent abstractions with Semantic Kernel's session state, type safety and telemetry. Graph-based workflows add explicit control over multi-agent execution. Microsoft provides migration guides for developers moving from either project.
Connect a model client, give an agent instructions and tools, then run it. Use sessions to maintain conversation state, context providers to supply memory, and middleware to intercept agent actions.
The Harness Agent adds planning, task tracking and context compaction for longer jobs. Use workflows when a process has defined steps or needs several agents and functions to coordinate. Open-ended conversations can use an agent on its own.
The Go SDK is in public preview and lacks RAG, declarative agents and functional workflows. Evaluate capabilities for the language you use. Developers remain responsible for testing reliability, managing data boundaries and implementing AI safeguards.
The free option is the open-source SDK. Choose model providers and hosting separately when planning operating costs.
It combines tool use, conversation state and explicit workflow paths to coordinate agents and functions across different model providers.
Developers use this open source SDK to build AI agents and workflows in Python and .NET, with tools, memory and MCP integration.