News

Microsoft Ships Production-Ready Agent Framework 1.0 for .NET and Python

• By David Ramel
• 04/06/2026

Microsoft has released version 1.0 of Microsoft Agent Framework, marking what the company calls the production-ready milestone for its open-source SDK and runtime for building AI agents and multi-agent workflows.

The release is available for both .NET and Python, with Microsoft describing it as the stable release with long-term support. The company is pitching the framework as a foundation for developers who want to move beyond simple assistants into orchestrated agent-based systems that can call tools, work across multiple models, and participate in longer-running workflows.

In practical terms, Microsoft Agent Framework is the company's attempt to provide one developer stack for building, orchestrating, and deploying AI agents and multi-agent applications. It combines work that previously lived in separate Microsoft efforts, especially Semantic Kernel and AutoGen, into a single open-source framework intended to support both experimentation and enterprise deployment.

Microsoft announced the new milestone in an April 3 post, saying: "This is the production-ready release: stable APIs, and a commitment to long-term support." The company added that Agent Framework 1.0 is designed to provide "enterprise-grade multi-agent orchestration, multi-provider model support, and cross-runtime interoperability via A2A and MCP."

That positioning follows the project's original debut last fall. In the earlier October 2025 introduction announcement, Microsoft described Agent Framework as "an open-source SDK and runtime" that "unifies the enterprise-ready foundations of Semantic Kernel with the innovative orchestration of AutoGen." At the time, the company framed the new offering as a way to give developers one foundation spanning open standards, research-to-production workflows, extensibility, and enterprise readiness.

Visual Studio Magazine covered that introduction in October in "Semantic Kernel + AutoGen = Open-Source 'Microsoft Agent Framework'," when Microsoft first laid out its plan to merge the strengths of the two projects into a unified framework.

Before Agent Framework, Microsoft's two main open-source building blocks in this area were Semantic Kernel and AutoGen. Semantic Kernel was the company's model-agnostic SDK for integrating AI capabilities into applications and for building and orchestrating agents and multi-agent systems, while AutoGen was focused on creating multi-agent AI applications in which agents could act autonomously or collaborate with humans. As for status, Microsoft now positions Agent Framework as the successor for new agent-development work; AutoGen is still being maintained with bug fixes and critical security patches, while Semantic Kernel remains active, with recent releases still appearing in its GitHub repo.

What Reached 1.0
For the 1.0 release, Microsoft said the stabilized surface includes the core single-agent abstraction and service connectors across .NET and Python; middleware hooks; agent memory and context providers; graph-based workflows; and multi-agent orchestration patterns including sequential, concurrent, handoff, group chat, and Magentic-One. Microsoft also said the 1.0 feature set carries a commitment to backward compatibility going forward.

Microsoft further highlighted support for multiple model and service providers. In the 1.0 announcement, the company said Agent Framework ships with first-party connectors for Microsoft Foundry, Azure OpenAI, OpenAI, Anthropic Claude, Amazon Bedrock, Google Gemini, and Ollama.

The company is also continuing its standards-based messaging around interoperability. In the original October announcement, Microsoft emphasized support for MCP (Model Context Protocol), A2A (Agent-to-Agent), and OpenAPI-oriented integration. In the 1.0 post, it reiterated MCP support and said A2A 1.0 support is coming soon, positioning those protocols as part of its cross-runtime agent story.

What Changed Since the October Introduction
Microsoft's April post draws a sharper line between what is now stable and what remains in preview. The company said the 1.0 release represents the features it has "battle-tested, stabilized, and committed to supporting," while also listing a set of newer capabilities still being shipped as preview features.

Those preview items include DevUI, described as a browser-based local debugger for inspecting agent execution and orchestration behavior in real time; Foundry hosted agent integration; Foundry-backed tools, memory, observability, and evaluations; front-end adapters for AG-UI, CopilotKit, and ChatKit; reusable "skills"; GitHub Copilot SDK and Claude Code SDK integration; and an "Agent Harness" for shell, file system, and messaging-loop access in coding and automation scenarios.

That split matters because Microsoft is now drawing a distinction between the production-ready core and the adjacent capabilities that are available for early adoption but may still evolve. For developers evaluating the platform, the message is that the basic agent, workflow, memory, middleware, and orchestration building blocks are now the supported center of gravity.

Microsoft had already detailed other work on the project and signaled that the framework was nearing completion in a February release-candidate announcement, when it said the API surface was stable for both .NET and Python and that all features planned for version 1.0 were complete. That set up the April 3 general-availability-style release as less of a feature reveal and more of a production-readiness milestone.

Open Source, Multi-Language, and Migration Path
The project's GitHub repository describes Microsoft Agent Framework as "A framework for building, orchestrating and deploying AI agents and multi-agent workflows with support for Python and .NET." The repository README also presents the framework as a multi-language project spanning simple agents through graph-based multi-agent workflows, with documentation, tutorials, quickstarts, and migration guidance for existing Semantic Kernel and AutoGen users.

Microsoft is also being careful to describe Agent Framework as an evolution, not a hard break. In the October introduction post, the company said the new framework "doesn't replace Semantic Kernel and AutoGen -- it builds on them." That theme continues in the 1.0 messaging, which points developers to migration guides for both predecessor projects and says now is the time to move from those codebases to Agent Framework.

For Microsoft-centric developers, the release is notable not only because it formalizes the company's agent-development story, but because it does so in a way that spans both .NET and Python while tying together previously separate Microsoft AI tooling efforts. With 1.0 now out, the question shifts from what the framework is supposed to become to how quickly developers will adopt it as the default foundation for production agent applications.