microsoft-agent-framework

Follow the official migration guide at learn.microsoft.com/agent-framework/migration-guide/from-semantic-kernel. Preserve existing behavior first, then adopt native Agent Framework patterns incrementally rather than rewriting everything at once.

For Python, run pip install agent-framework. For .NET, run dotnet add package Microsoft.Agents.AI. Check the official repository samples for the latest package versions since the framework is in public preview.

Use agents for autonomous decision-making, ad hoc planning, conversation flows, tool usage, and MCP server interactions. Use workflows for multi-step orchestration, predefined execution graphs, long-running tasks, and human-in-the-loop scenarios.

Yes, Microsoft Agent Framework is the unified successor to Semantic Kernel and AutoGen, combining their strengths with new capabilities. Older patterns from those frameworks should be treated as migration inputs, not the default implementation model.

It supports Azure AI Foundry, Azure OpenAI, OpenAI, and other providers. Azure AI Foundry is preferred for new projects when it matches user needs, and DefaultAzureCredential is recommended when Azure authentication is appropriate.