Getting Started: Model Context Protocol (MCP)

Learn the fundamentals of MCP, the open standard for connecting AI assistants to external tools, data sources, and APIs.

About This Course

Most developers are still glueing AI applications together with brittle prompt engineering and fragile workarounds. This course teaches you how to move beyond that by designing and shipping production-ready AI applications using the Model Context Protocol (MCP) in C#. You'll build both an MCP Server and an MCP Client from scratch, connect them over StreamableHTTP, and integrate them into a working AI agent powered by OpenAI. Along the way, you'll tackle real-world concerns like authentication with Microsoft Entra ID, secure tool design, trust boundaries, resource exposure, orchestration, and deployment to Azure. By the end, you won't just understand the MCP spec. You'll know how to use it to build secure, maintainable, distributed AI systems that are ready for production.

Course Curriculum

6h 11m 9 sections

Welcome

02:07 Free preview

What will you learn in this course?

02:21 Free preview

Who is the course for and prerequisites

01:54 Free preview

The Problem With Prompt-Centric Systems

06:31

MCP in Distributed AI Systems

04:51

What Is An MCP Server

04:55

Local vs Remote MCP Servers

09:04

What MCP Doesn’t Solve

04:30

MCP Primitives (Tools, Prompts, Resources)

03:19

Hosts, Clients and Servers

05:01

The Request & Response Flow

05:37

MCP Lifecycle

04:19

The StreamableHTTP Transport

04:45

MCP Helper Libraries for .NET

02:04

Functional Testing in MCP Inspector

05:32

C# MCP Server with stdio

09:48

C# MCP Server with Streamable HTTP

09:08

Adding Database MCP Tools

23:33

Tools vs Skills in Modern Agents

06:25

What an MCP Client Is Responsible For

02:21

When to Build Your Own Client vs Reuse One

02:41

Agent Design with Microsoft Agent Framework

08:03

Creating a Basic Chat Agent in .NET

32:11

Connecting Your MCP Server

29:23

Context vs State vs Memory

08:37

Why Security Is Harder with MCP

07:46

Defining Trust Boundaries in MCP Systems

07:50

Adding Entra ID to the MCP Server

17:05

Creating Tokens & Testing Authentication

08:45

Read vs Write Tools

06:58

Setting Up Entra ID For OBO Flow

09:21

Adding Authentication To The MCP Client

26:38

Making Model Friendly Tools

05:14

Schema Design That Models Handle Well

06:06

Naming, Discoverability & Documentation

09:38

Adding some more tools into the server

12:57

What MCP Resources Are For

02:24

Choosing Resources vs Tools

04:56

Dynamic and Computed Resources

02:47

Adding MCP Resources into the server

14:05

From Local Tool to Production Service

01:40

Deployment Models for MCP Servers

05:44

Deploying To Azure With Docker

13:58

Distribution Channels for MCP Servers

07:25

Course Details

Level Getting Started

Duration

$119.99

Lifetime access

Go Pro for Full Access

Get unlimited access to all Dometrain courses with a Pro subscription

Learn more about Dometrain Pro

Meet Your Instructor

James Charlesworth

James Charlesworth

James is a Director of Engineering, O'Reilly author, speaker, and YouTuber with 20 years of experience in the tech industry. He has worked across startups and publicly listed tech companies, helping to build and scale software products. James focuses on teaching product skills to engineers, building high-performing product engineering teams, and defining the "Product Engineer" career path. He shares practical advice on software engineering, product development, and career growth through his Train to Code YouTube channel.

View all courses by James Charlesworth

© 2026 Dometrain. All rights reserved.