MCP Winter Nights
Join Mihaela, Andrei, Daniel and Vlad for an evening diving into how MCP reshapes software integration end to end.
From rethinking traditional APIs for AI consumers, to building and comparing MCP-powered autonomous agents, to implementing production-ready MCP servers in .NET 10, you’ll get a practical, code-first tour of anything and everything MCP.
Who
Mihaela Ghidersa-Pintilei, Microsoft MVP on Web Development, .NET
Andrei Ignat, Microsoft MVP on .NET, Web Development
Daniel Costea, Microsoft MVP on Azure AI Foundry, .NET
Vlad Iliescu, Microsoft MVP on Azure AI Foundry
Agenda
18:20 - 18:30 Welcome
18:30 - 19:00 Mihaela - MCP vs Traditional APIs
19:00 - 19:30 Vlad - MCP vs Tools for Autonomous Agents
19:30 - 20:00 Andrei - MCP and .NET 10
20:00 - 20:30 Daniel - Microsoft Agentic Framework and MCP
Talk Descriptions
MCP vs Traditional APIs: How Software Integration Changes When AI Is the Consumer (Mihaela)
For years we’ve built APIs for human developers — REST, GraphQL, gRPC — all assuming the consumer reads documentation, understands endpoints, and writes code manually.
With MCP, the consumer is different: it’s an AI. In this talk, we’ll explore how MCP enables AI agents to discover, understand, and execute application capabilities dynamically.
We’ll compare traditional API integrations with MCP-style capability exposure and look at practical examples of how this changes the way we build and expose services.
MCP vs Tools for Autonomous Agents (Vlad)
We'll see what it means to implement from scratch an MCP server able to connect to Microsoft Foundry, and discuss the pros and cons of using said MCP server when building autonomous agents.
We'll also discuss alternate, simpler ways of building rich agents, all the while building one such agent ourselves using Python, OpenAI, and a little library called ZipCoil.
MCP and .NET 10 (Andrei)
This session will walk you through designing and building a Model Context Protocol (MCP) server with .NET 10, starting from a clean solution. You will see how to structure the project, implement practical MCP tools and resources, and use new .NET 10 features to keep the code simple and maintainable.
We will then cover how to package, deploy, and connect the MCP to an AI client in a real environment. By the end, you will have a clear end-to-end picture: from first line of code to a working .NET 10–based MCP.
Microsoft Agentic Framework and MCP (Daniel)
The tools are a core element of any AI agent, and in .NET almost any traditional function can be exposed as a tool. However, when we want to standardize these tools, we need a way to decouple them from the application code. This is where MCP fits: it moves tools into a dedicated MCP service, which any application can then consume through an MCP client. In this talk we'll see how we can convert and host tools as MCP tools and consume them with our app using Microsoft Agent Framework.