Build Your First AI Agent in .NET

Come spend two days with me and we'll build a personal finance assistant step by step. You'll ask it questions, give it tools, let it work with your data and let it take actions for you.

Day 1: Build the agent

• We'll start with the simplest possible version: calling an LLM over HTTP.
• From there, we'll look at how prompts and tokens work before adding abstractions.
• You'll give the assistant tools to read data and take action, and build the loop that decides when those tools should be called.
• Then we'll add agentic RAG so it can find transactions by meaning, not just exact matches.
• We'll also keep conversation history, so you can ask follow-up questions without having to repeat everything.

Day 2: Prepare it for production and connect it to other tools

• Long conversations eventually become a problem, so we'll summarise older messages and manage the context we send to the model.
• We'll trace what the agent is doing, look at prompt injection, and add confirmation before actions that could have consequences.
• Then we'll write evaluations so you can change prompts, tools or models and still check whether the agent behaves the way you expect.
• You'll also build your own MCP server and use it from Claude Desktop, VS Code or another MCP client.
• Finally, we'll take the agent we've built and see what it looks like using Microsoft Agent Framework, including where a framework helps and where you may not need one.

During the workshop, we'll use Microsoft Foundry. If you prefer, you can bring your own OpenAI API key. The workshop project makes it easy to switch between the two.

What you'll leave with

• A working agent built with C# and .NET.
• An MCP server you've built yourself.
• An understanding of the pieces behind tools like Codex, Copilot and Claude.
• And, most importantly, a project you can take home, change and use as a starting point when you add AI to your own applications.

Who is this for?

If you answer yes to these two questions, you should feel comfortable in the workshop:

• Are you comfortable building APIs in C#?
• Do you prefer learning by building things instead of sitting through slides and theory?

What you'll need

• Your favourite IDE (Visual Studio 2026, if you use Visual Studio)
• .NET 10
• Docker
• Git and a GitHub account
• An Azure account or an OpenAI API key. Expect up to 5€ of usage.