AI agents are quickly becoming a first-class part of modern applications, and two open protocols are defining how they connect. The Model Context Protocol (MCP) has become the standard way to give an agent tools, data, and enterprise capabilities. The Agent2Agent (A2A) protocol lets agents built on different frameworks, by different teams, discover each other and delegate work. In this session, you’ll learn how to build both with C# and .NET, so your business logic, APIs, and data are available to individual agents and to the multi-agent systems they work in.

We’ll walk through the fundamentals of MCP and A2A, starting with how to implement MCP tools and resources in .NET, design servers for the new stateless MCP spec so they scale like any other web workload, and integrate with VS Code, OpenAI clients, and Microsoft Foundry Agent Service. Then we’ll move up a layer to A2A. You’ll publish an agent card so other agents can find your agent, handle long-running tasks, and orchestrate specialist agents that each use their own MCP tools. You’ll see end-to-end examples, from simple utilities to enterprise workflows that connect to SQL, Azure services, Microsoft Graph, and more.

Whether you’re exposing your .NET systems as reusable tools or building agents that collaborate across teams and platforms, this session will show you how to turn your .NET expertise into powerful AI integrations. You’ll leave with a clear rule for when to build an MCP tool and when to build an A2A agent, plus the code to build either one.

You will learn:

• About the Model Context Protocol (MCP) and the Agent2Agent (A2A) protocol
• How to build MCP servers and A2A agents using C# and .NET
• How to design, develop, and deploy enterprise-ready MCP servers and A2A agents