Microsoft Agent Framework Hits 1.0, Merging AutoGen and Semantic Kernel
Microsoft unifies Semantic Kernel and AutoGen into Agent Framework 1.0 with stable APIs, multi-agent workflows, and native A2A/MCP support.
Microsoft shipped Agent Framework 1.0 this week, and for once the version number means something. The release unifies Semantic Kernel's enterprise engineering — the stuff that's been quietly powering Copilot integrations inside Fortune 500 companies — with AutoGen's orchestration layer, the framework that made multi-agent conversations approachable for the rest of us. If you've been watching this space, you know that split caused real confusion. AutoGen handled multi-agent workflows beautifully but had gaps in enterprise-grade concerns like Observability and policy enforcement. Semantic Kernel had the opposite problem: robust, enterprise-ready, but verbose and complex for anything beyond single-agent patterns. Version 1.0 is Microsoft's answer to the question everyone was asking: why can't we have both?
The technical stack delivers on that promise with stable, non-breaking APIs — finally — alongside native support for both the A2A and MCP protocols that the industry has been converging on. Cross-provider model support is built in, so you're not locked into Azure OpenAI or the Microsoft stack. The getting-started experience is what caught my eye: the team is advertising "from zero to agent in five lines of code," and from a quick scan of the docs, they're not lying. That's not nothing. One of the biggest friction points for enterprise teams evaluating agent frameworks has been the gap between the hello-world tutorials and production readiness. Five lines to a working agent removes one excuse for that gap.
Under the hood, the framework ships with multi-agent workflow primitives that handle the coordination patterns you'd expect: task delegation, result aggregation, shared state across agent boundaries. The .NET and Python SDKs are both at 1.0, which matters for polyglot shops. Microsoft is positioning this as the recommended path for anyone building on the Azure AI infrastructure, but the cross-provider story means it's a legitimate option for teams running on AWS or GCP too. The semantic-kernel-to-agent-framework migration path is documented, which suggests the team expects Semantic Kernel users to upgrade rather than maintain two codebases.
What does this mean for the broader agent framework landscape? AutoGen was effectively sidelined into maintenance mode — Microsoft's own docs now point users to the unified framework. LangGraph is still the default for teams that want graph-based workflow definition and don't need the enterprise integrations. But if you're already in the Microsoft ecosystem, or if you need the policy controls and compliance story that Semantic Kernel provided, this is a meaningfully better starting point than either predecessor. The fragmentation problem in agent frameworks is real; this release doesn't solve it, but it does reduce the number of Microsoft-branded options from two to one, which is progress.