
Microsoft's open-source agent framework for Python and .NET, the 1.0 successor to both Semantic Kernel and AutoGen, with graph-based workflows and hosting on Foundry Agent Service.
Microsoft Agent Framework is where Microsoft's two earlier agent projects ended up. Its documentation calls it "the direct successor, created by the same teams" and "the next generation of both Semantic Kernel and AutoGen": it keeps AutoGen's simple single- and multi-agent abstractions, adds Semantic Kernel's enterprise features (session-based state, type safety, middleware, telemetry), and introduces graph-based workflows that give explicit control over multi-agent execution paths, with state management for long-running and human-in-the-loop work. Version 1.0 was announced for Python and .NET on 3 April 2026 and both SDKs release frequently (python-1.25.0 and dotnet-1.28.0 in the first week of November 2026). A Go SDK is in public preview in a separate repository and does not yet implement declarative agents or functional workflows. Model support covers Microsoft Foundry, Azure OpenAI, OpenAI, Anthropic and Ollama among others, and agents can call MCP servers. Microsoft publishes migration guides from both Semantic Kernel and AutoGen, which matters because Microsoft has declared AutoGen archived and no longer maintained. The framework is free; the cost question is where you run it. Microsoft's own advice applies before you start: "If you can write a function to handle the task, do that instead of using an AI agent."
Get a new AI workflow each week — many feature Microsoft Agent Framework and other tools in this category.