MCP with .NET: securely exposing your data to LLMs

Model Context Protocol is rapidly becoming the de facto standard for LLMs and systems to communicate.

This session explores the fundamentals of the MCP protocol, how to safely implement it into our applications with the .NET SDK and Microsoft Agent Framework / Semantic Kernel, plus key considerations for building a production-ready system – like security, auth, rate limits, and observability.

We'll cover the basics of setting up an MCP server in .NET and exposing executable functionality as Tools, enabling any client to communicate with our application. We'll understand how Semantic Kernel's familiar concepts of Plugins and Functions map over to the world of MCP.

Let's walk through integrating .NET's native Identity services to align with MCP’s authentication spec, enforce role-based permissions using Filters, and apply rate limiting middleware to keep the server reliable and secure.

Finally we will discuss approaches to running intelligent .NET applications in the wild on Azure; with .NET Aspire providing scaling and OpenTelemetry, deployed on Azure Container Apps, and using an AI Gateway with appropriate policies via Azure API Management to create a hardened system.