Microsoft shipped Agent Framework v1.0 on April 3, completing a six-month consolidation of two previously incompatible open-source projects: AutoGen, the multi-agent research framework from Microsoft Research, and Semantic Kernel, the enterprise-grade orchestration SDK. The combined release is production-stable, carries a backward-compatibility commitment, and ships for both .NET and Python. According to Forbes analysis published Monday, developers are finding the unified stack confusing while rivals like LangChain and OpenClaw continue to gain ground by making more opinionated choices for builders.

What Ships in v1.0

Per the official v1.0 announcement:

  • A stable agent abstraction across .NET and Python with a long-term support commitment
  • First-party service connectors for Microsoft Foundry, Azure OpenAI, OpenAI, Anthropic Claude, Amazon Bedrock, Google Gemini, and Ollama
  • Sequential and concurrent multi-agent workflow orchestration
  • Cross-runtime interoperability via both A2A (Agent-to-Agent Protocol) and MCP (Model Context Protocol)
  • A middleware pipeline for content safety filters, logging, and compliance policies

Microsoft’s pitch: “from zero to agent in 5 lines of code.” The framework was announced in October 2025 as a public preview, hit Release Candidate in February 2026, and now carries a GA label after six months of community validation with customers and partners.

The Consolidation Problem

The original rationale for merging AutoGen and Semantic Kernel was straightforward: two teams inside Microsoft had built two frameworks that solved overlapping problems but couldn’t be used together cleanly. AutoGen brought cutting-edge research on multi-agent coordination. Semantic Kernel brought production-hardened tooling suitable for enterprise IT. The merge would give developers both without having to choose.

According to Forbes, the result hasn’t delivered on that simplicity promise. Individual developers and smaller teams have been gravitating toward more opinionated tools that make explicit decisions about agent structure and execution flow. LangChain’s three-layer framework and OpenClaw’s skill-based architecture both offer clearer onboarding paths. Microsoft’s 1.0 release inherits the tension between enterprise configurability and research-grade flexibility, producing a framework with significant breadth that still makes fewer decisions for you than its competitors.

The Azure Distribution Angle

Where Microsoft has an obvious structural advantage is distribution. Agent Framework 1.0 is built to slot into Azure AI Foundry project configurations, work with Microsoft’s existing enterprise tooling, and ship inside the same compliance envelopes that Fortune 500 IT departments already use. Multi-provider model support, including connectors for Anthropic and Amazon Bedrock, reduces lock-in risk for teams that want to hedge model providers.

For teams running inside the Microsoft stack who want enterprise-grade agent orchestration with a long-term support commitment, v1.0 is the clearest path forward Microsoft has offered. For teams building new in 2026 without Azure ties, the question is whether the breadth of the framework justifies navigating its complexity over a lighter, more opinionated alternative.