Microsoft just hit a significant milestone. On April 3, the company announced version 1.0 of its Microsoft Agent Framework — a production-ready, open-source SDK for building, orchestrating, and deploying AI agents. It’s available for both .NET and Python, with stable APIs and a long-term support commitment.
This isn’t a preview or a proof of concept. It’s the real thing.
One Framework to Replace Two
If you’ve been following Microsoft’s AI development tooling, you know the company has been running two separate frameworks: Semantic Kernel and AutoGen. Each had its strengths — Semantic Kernel for enterprise-ready stability, AutoGen for more experimental orchestration patterns. Agent Framework 1.0 is the unified successor to both.
The framework is open source and designed to build, orchestrate, and deploy AI agents with a consistent programming model across .NET and Python. Both Semantic Kernel and AutoGen are now in maintenance mode — security patches and bug fixes will continue, but new feature development is moving to Agent Framework going forward.
For development teams still on either of those frameworks, migration guides are included in the 1.0 release.
Getting Started Is Straightforward
One of the more practical aspects of this release is how quickly you can stand up a working agent. The framework takes you from zero to agent in five lines of code, in either language. From there, you can layer in function tools, multi-turn sessions, streaming responses, and more complex orchestration.
What’s in the Box
The 1.0 release is built around a set of core capabilities that have been tested, stabilized, and committed to backward compatibility:
Single-agent and service connectors are stable and production-ready for both .NET and Python. Agent Framework ships with first-party connectors for Microsoft Foundry, Azure OpenAI, Anthropic Claude, Amazon Bedrock, Google Gemini, and Ollama. That’s a notable list — this isn’t a Microsoft-only ecosystem.
A middleware pipeline lets developers intercept, transform, and extend agent behavior at every stage of execution — content safety filters, logging, compliance policies, and custom logic — without modifying agent prompts.
Memory is pluggable. The framework supports conversational history, persistent key-value state, and vector-based retrieval, with backend options including Memory in Foundry Agent Service, Mem0, Redis, and Neo4j.
For teams building more complex systems, the graph-based workflow engine lets developers compose agents and functions into deterministic, repeatable processes. Workflows can combine agent reasoning with business logic, branch on conditions, fan out to parallel steps, and converge results. Checkpointing and hydration ensure long-running processes survive interruptions.
Multi-agent orchestration patterns from Microsoft Research and AutoGen are also included — sequential, concurrent, handoff, group chat, and Magentic-One — all with support for streaming, checkpointing, and human-in-the-loop.
Interoperability is Built In
Two protocol integrations stand out here. MCP (Model Context Protocol) enables agents to dynamically discover and invoke external tools exposed over MCP-compliant servers. A2A (Agent-to-Agent) protocol support enables cross-runtime agent collaboration — agents can coordinate with agents running in other frameworks using structured, protocol-driven messaging.
For enterprises concerned about vendor lock-in, this matters. A single MCP server can be shared across agents in different frameworks from different vendors.
Developer Experience
The 1.0 release also ships a browser-based local debugger called DevUI — a tool for visualizing agent execution, message flows, tool calls, and orchestration decisions in real time. It’s in preview, but it addresses a real need: Debugging multi-agent systems has historically been difficult.
Other preview features include deep integration with Microsoft Foundry for managed hosting, adapters for frontend surfaces such as CopilotKit and ChatKit, and experimental support for the GitHub Copilot SDK and Claude Code SDK as agent harnesses within multi-agent workflows.
What This Means for Enterprise Teams
Teams can now move from prototype to production in hours rather than weeks, building and testing multi-agent AI systems with enterprise-grade security — without Kubernetes or container complexity. That’s a practical reduction in friction for teams that want to ship production AI systems without becoming infrastructure specialists.
The framework also supports declarative agent and workflow definition in YAML — a feature that makes version control and configuration management significantly easier.
Mitch Ashley, VP and practice lead for software lifecycle engineering at The Futurum Group, sees the release as a meaningful step forward for enterprise AI development. “Microsoft’s consolidation of Semantic Kernel and AutoGen into a single production-ready foundation reflects vendors competing to own the enterprise agent orchestration layer. Multi-provider connectors for Anthropic Claude, Amazon Bedrock, Google Gemini, and others, combined with MCP and A2A protocol support, position Agent Framework 1.0 as infrastructure for multi-vendor, multi-cloud agent deployments.”
“The middleware pipeline for compliance, logging, and policy enforcement delivers the governance hooks production deployments require. Teams still on Semantic Kernel or AutoGen have documented migration paths and a clear direction forward; teams building new agentic systems have a stable, extensible foundation and no reason to defer.”
The Bottom Line
Microsoft Agent Framework 1.0 is a well-considered consolidation of two capable but fragmented developer tools into a single, production-grade foundation. The multi-provider model supports open protocol integrations via MCP and A2A, and a pluggable memory architecture gives enterprise teams genuine flexibility. And the “zero to agent in five lines” promise isn’t marketing — the code samples back it up.
If you’re building agentic AI systems on Microsoft infrastructure — or even if you’re not — this framework is worth a close look.
The SDK is available now at aka.ms/AgentFramework.
