You start a new AI feature in .NET, search for "Semantic Kernel", and copy the first sample you find. Six months later the SDK only gets security fixes, and every new capability you need lives somewhere else.
Semantic Kernel and AutoGen are in maintenance mode. Microsoft Agent Framework is their successor, built by the same teams.
Microsoft Agent Framework is Microsoft's open-source SDK for building AI agents and multi-agent workflows in .NET and Python. It combines the simple agent abstractions of AutoGen with the enterprise features of Semantic Kernel, and it reached 1.0 with stable APIs and long-term support.
It gives you:
using Azure.AI.Projects;using Azure.Identity;using Microsoft.Agents.AI;AIAgent agent = new AIProjectClient(new Uri("https://your-foundry-service.services.ai.azure.com/api/projects/your-foundry-project"),new AzureCliCredential()).AsAIAgent(model: "gpt-5.4-mini",instructions: "You are a friendly assistant. Keep your answers brief.");Console.WriteLine(await agent.RunAsync("What is the largest city in France?"));
A new project that references Microsoft.SemanticKernel and builds a Kernel to call the model.
❌ Figure: Bad example - Bad example - Starting new work on an SDK in maintenance mode
A new project that references Microsoft.Agents.AI and creates an AIAgent from the model client.
✅ Figure: Good example - Good example - Starting new work on Microsoft Agent Framework
| Use an agent when... | Use a workflow when... |
|---|---|
| The task is open-ended or conversational | The process has well-defined steps |
| The model should plan and pick its own tools | You need explicit control over execution order |
| A single model call (with tools) is enough | Several agents or functions must coordinate |
If a plain function can do the job, write the function. An agent adds cost and non-determinism.
Existing Semantic Kernel and AutoGen apps keep working, so there is no need to rewrite them today. Move them when you next make a significant change, using the official guides:
For how Microsoft Agent Framework compares with LangGraph, the OpenAI Agents SDK, and n8n, see Do you build agentic AI?