Module 2 of 2 in AI Agents from Scratch

MCP Fundamentals

Apr 10 2026, Python 3, JavaScript, macOS , Windows, VS Code
Leave a rating/review

Learn how to build real applications with the Model Context Protocol (MCP), from first principles to advanced integrations. In this course, you will start by understanding the core MCP architecture and building your first server with Python and FastMCP. You will then add Tools, Resources, and Prompts, inspect them with MCP Inspector, and move on to building custom MCP clients that can work programmatically with LLMs through the Anthropic API. From there, you will explore advanced features such as Elicitation for human-in-the-loop workflows, Roots for filesystem security, and Sampling for client-side AI execution. Finally, you will bring everything together by building a full-stack ChatGPT App that serves a React frontend from a Python MCP backend using the OpenAI Apps SDK. By the end of the course, you will understand how MCP hosts, clients, and servers fit together, how to design reliable tool schemas and resources, and how to ship MCP-powered experiences that work in desktop clients, custom programs, and ChatGPT. By arjuna sky kok.

Start learning for free
 
EXPERIENCE
Intermediate level
EST TIME TO COMPLETE
2 hrs, 11 mins
CONTENT
5 lessons
Share
Save for later

Module outcomes

  • Build MCP servers with Python and FastMCP
  • Implement and inspect MCP Tools, Resources, and Prompts
  • Create programmatic MCP clients and connect them to LLM APIs
  • Apply advanced MCP features such as Elicitation, Roots, and Sampling
  • Ship a full-stack ChatGPT App with a Python backend and React frontend

Covered concepts

  • MCP architecture: Hosts, Clients, and Servers
  • FastMCP server development in Python
  • Tools, Resources, and Prompts
  • MCP Inspector and schema validation
  • Programmatic MCP clients and Anthropic MCP Connector
  • Elicitation, Roots, and Sampling
  • OpenAI Apps SDK and ChatGPT widgets
  • Serving React frontends as MCP resources

Module content

1
What is MCP & Hello World Lesson (18 mins)

This lesson introduces students to the Model Context Protocol (MCP), a standardized approach to connecting AI models with external tools and data. Students will explore the architecture of MCP, understanding the roles of Hosts, Clients, and Servers. The lesson covers setting up a development environment, writing a functional MCP server using Python and FastMCP, and successfully integrating it with Claude Desktop.

 Start
Introduction • Start Reading · 1 min
Overview of Model Context Protocol Reading · 6 mins
Setting Up Development Environment Video · 2 mins
Writing MCP Server Code Reading · 3 mins
Running MCP Server Video · 4 mins
Conclusion Reading · 1 min
2
Tools, Resources, Prompts, and Model Inspector Lesson (29 mins)

This lesson dives deep into the three core building blocks of the Model Context Protocol: Tools, Resources, and Prompts. You will learn how to implement each component using Python and the FastMCP framework, and how to verify their functionality using the MCP Inspector.

 Start
Introduction • Start Reading · 1 min
Setting Up MCP Inspector Video · 2 mins
Understanding MCP Tools Reading · 7 mins
Executing Tools with Inspector Video · 3 mins
Understanding MCP Resources Reading · 6 mins
Inspecting Resources with Inspector Video · 3 mins
Understanding MCP Prompts Reading · 4 mins
Inspecting Prompts with Inspector Video · 2 mins
Conclusion Reading · 1 min
3
Building MCP Clients & LLM Integration Lesson (24 mins)

In this lesson, you will transition from using graphical interfaces to building programmatic MCP clients. You will learn how to write Python scripts that directly interact with MCP servers and how to integrate Large Language Models via API to create autonomous, tool-using agents.

 Start
Introduction • Start Reading · 2 mins
Building an MCP Client Reading · 8 mins
Running the MCP Client Video · 3 mins
Programmatic MCP with LLMs Reading · 7 mins
Connecting Claude to Local Tools Video · 4 mins
Conclusion Reading · 1 min
4
Advanced Features Lesson (33 mins)

In this lesson, you will master the advanced capabilities of the Model Context Protocol. You will learn how to implement “Human-in-the-loop” workflows with Elicitation, secure your filesystem with Roots, and leverage client-side intelligence via Sampling.

 Start
Introduction • Start Reading · 2 mins
Understanding Elicitation Reading · 10 mins
Handling Elicitation with a Custom Client Video · 2 mins
Understanding Roots Reading · 8 mins
Enforcing Roots Video · 2 mins
Understanding Sampling Reading · 6 mins
Sampling with Real AI Video · 3 mins
Conclusion Reading · 1 min
5
Building a ChatGPT App Lesson (27 mins)

In this lesson, you will go beyond text-based interactions and build a full-stack graphical application that lives inside ChatGPT. You will learn how to serve a React frontend from your Python MCP backend, use the OpenAI Apps SDK to bridge the two, and deploy the result as an interactive widget.

 Start
Introduction to ChatGPT Apps • Start Reading · 3 mins
The Backend: Serving Resources Reading · 10 mins
The Frontend: The OpenAI Bridge Reading · 5 mins
Server & Tunnel Setup Video · 3 mins
Running the Application Video · 5 mins
Conclusion Reading · 1 min