Microsoft Agent Framework 1.0 Ships with Stable .NET and Python APIs

Why It Matters for Developers#

For teams that were waiting for a stable release before committing, the calculus changed. The core API surface is tracked in shipped public API files in the repository, a common discipline for stable .NET libraries that makes accidental breaking changes visible in code review. With the current packages, a tool-using agent takes only a few lines:

using System.ComponentModel;
using Microsoft.Agents.AI;
using Microsoft.Extensions.AI;
using OpenAI.Chat;

var apiKey = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;
IChatClient chatClient = new ChatClient("gpt-4o-mini", apiKey).AsIChatClient();

AIAgent agent = chatClient.AsAIAgent(
    instructions: "You answer questions about order status. Be concise.",
    name: "OrderAgent",
    tools: [AIFunctionFactory.Create(GetOrderStatus)]);

AgentSession session = await agent.CreateSessionAsync();
AgentResponse response = await agent.RunAsync("Where is order 1042?", session);
Console.WriteLine(response.Text);

[Description("Gets the shipping status for an order.")]
static string GetOrderStatus([Description("The order number.")] int orderId) =>
    orderId == 1042 ? "Shipped on April 1" : "Unknown order";

A few practical recommendations follow from how the release is structured:

• Build on the stable core, isolate the rest. Agents, workflows and the OpenAI and Foundry integrations are stable. If you depend on preview packages such as A2A or the Anthropic integration, pin their versions and wrap them behind your own interfaces.
• Treat experimental orchestrations as experimental. Handoff patterns can change in minor releases. Sequential and concurrent workflows are a safer default for business-critical processes.
• Plan Semantic Kernel migrations concept by concept. Microsoft's migration guide maps [KernelFunction] plugins to tools created with AIFunctionFactory, the various Semantic Kernel agent types to a single ChatClientAgent, threads to AgentSession objects created by the agent, and InvokeAsync to RunAsync. Our Semantic Kernel guide helps you decide which parts to move first.
• Remember legacy reach. Because the packages still target .NET Framework 4.7.2 and .NET Standard 2.0, agents can be added to older line-of-business applications without waiting for a full platform migration.

For design guidance beyond the API, including when to use a single agent, a workflow or a multi-agent system, see AI agent architecture patterns for .NET and our hands-on Microsoft Agent Framework guide.