# Microsoft Agent Framework > Source: https://aiwiki.ai/wiki/microsoft_agent_framework > Updated: 2026-07-07 > Categories: AI Agents, Developer Tools, Microsoft > License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) > From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution to "AI Wiki (aiwiki.ai)". Microsoft Agent Framework (MAF) is an open-source SDK and runtime for building AI agents and multi-agent workflows in .NET and Python.[1][2] Developed by [Microsoft](/wiki/microsoft), it unifies the company's two earlier agent projects, the [AutoGen](/wiki/autogen) research framework and the [Semantic Kernel](/wiki/semantic_kernel) enterprise SDK, into a single platform.[2] Microsoft first released it in public preview in October 2025, brought it to release candidate on February 19, 2026, and shipped version 1.0 (general availability) on April 3, 2026.[1][5][3] The project is distributed under the MIT license.[4] ## What is Microsoft Agent Framework? Microsoft Agent Framework is a framework for building, orchestrating, and deploying [AI agents](/wiki/ai_agents) and multi-agent systems.[2][4] Its source repository describes it as "an open, multi-language framework for building production-grade AI agents and multi-agent workflows in .NET and Python."[4] The framework groups its capabilities into two primary categories. Individual agents use large language models to process input, call tools and Model Context Protocol (MCP) servers, and generate responses. Workflows are graph-based structures that connect agents and functions for multi-step tasks with type-safe routing, checkpointing, and human-in-the-loop support.[2] Microsoft guides developers to use an agent when a task is open-ended or conversational and to use a workflow when a process has well-defined steps that must run in a specific order.[2] Microsoft positions the framework for teams taking agents from prototype to production, providing a consistent foundation across the two languages while supporting a broad model ecosystem.[3] ## How does it relate to AutoGen and Semantic Kernel? The framework is the direct successor to both AutoGen and Semantic Kernel, created by the same engineering teams.[2] AutoGen began as a Microsoft Research project focused on simple abstractions for single- and multi-agent patterns, while Semantic Kernel supplied enterprise-grade features such as session-based state management, type safety, filters, telemetry, and broad model support.[1][2] Microsoft Agent Framework combines AutoGen's agent abstractions with Semantic Kernel's enterprise foundations and adds a graph-based workflow engine for explicit multi-agent orchestration.[2] Microsoft describes it as "the next generation of both Semantic Kernel and AutoGen."[2] ## What happened to AutoGen and Semantic Kernel? After the merger, both predecessor projects were retired in favor of Agent Framework. The AutoGen repository states that "AutoGen is now in maintenance mode. It will not receive new features or enhancements and is community managed going forward," and it directs new users to start with Microsoft Agent Framework.[6] The Semantic Kernel repository similarly announces that "Semantic Kernel is now Microsoft Agent Framework," positioning MAF as its enterprise-ready successor, and both projects supply migration guides.[2][7] Separately, a community-driven fork of AutoGen named AG2 continues independently of this transition. AG2 was established in late 2024 by original AutoGen contributors who had left Microsoft, and it is maintained as a community-governed project under the Apache 2.0 license, distinct from Microsoft's own AutoGen repository.[8] ## What languages and license does it use? Microsoft Agent Framework provides first-class support for both Python and C#/.NET, with consistent APIs across the two languages.[3][4] Python developers install it through the agent-framework package, while .NET developers use NuGet packages under the Microsoft.Agents.AI namespace.[2] The source code is released as open source under the MIT license on GitHub.[4] ## What are its main features? Beyond agents and workflows, the framework provides several building blocks:[2][3][4] - Orchestration patterns for coordinating multiple agents, including sequential, concurrent, handoff, group chat, and Magentic-One styles.[3] - Pluggable memory through context providers, supporting conversational history, persistent key-value state, and vector-based retrieval.[3] - Middleware hooks for intercepting, transforming, and extending agent behavior at each stage of execution.[3][4] - Model connectors for many providers, including Microsoft Foundry, Azure OpenAI, OpenAI, Anthropic Claude, Amazon Bedrock, Google Gemini, and Ollama.[3] - Interoperability standards, connecting to tools with MCP, integrating APIs through OpenAPI, and communicating with other agents using the Agent2Agent (A2A) protocol.[1] - Observability built on OpenTelemetry, plus a browser-based debugger called DevUI for visualizing agent execution.[3][4] The framework integrates tightly with Microsoft's Azure AI Foundry, including hosted agents, managed tools, and observability dashboards, although it can also run against non-Microsoft models and services.[2][3] ## How does it compare to other agent frameworks? Microsoft Agent Framework competes in a crowded field of agent-building toolkits that includes [LangChain](/wiki/langchain) and its graph-oriented counterpart [LangGraph](/wiki/langgraph), CrewAI, and OpenAI's Agents SDK. Its distinguishing traits are native, parallel support for both .NET and Python, deep integration with Azure AI Foundry, and a built-in graph-based workflow engine that pairs autonomous agents with deterministic, explicitly controlled execution paths.[2][3] Because it consolidates two widely used Microsoft frameworks and commits to stable APIs and long-term support at version 1.0, it is aimed particularly at enterprise teams already working within the Microsoft and Azure ecosystem.[3] ## References [1] "Introducing Microsoft Agent Framework." Microsoft Azure Blog, October 2025. https://azure.microsoft.com/en-us/blog/introducing-microsoft-agent-framework/ [2] "Microsoft Agent Framework Overview." Microsoft Learn. https://learn.microsoft.com/en-us/agent-framework/overview/ [3] "Microsoft Agent Framework Version 1.0." Microsoft Agent Framework Developer Blog, April 3, 2026. https://devblogs.microsoft.com/agent-framework/microsoft-agent-framework-version-1-0/ [4] "microsoft/agent-framework." GitHub. https://github.com/microsoft/agent-framework [5] "Migrate your Semantic Kernel and AutoGen projects to Microsoft Agent Framework Release Candidate." Microsoft Developer Blog, February 19, 2026. https://devblogs.microsoft.com/agent-framework/migrate-your-semantic-kernel-and-autogen-projects-to-microsoft-agent-framework-release-candidate/ [6] "microsoft/autogen." GitHub. https://github.com/microsoft/autogen [7] "microsoft/semantic-kernel." GitHub. https://github.com/microsoft/semantic-kernel [8] "ag2ai/ag2." GitHub. https://github.com/ag2ai/ag2