MCP: Build Rich-Context AI Apps with Anthropic

This repo covers how to use the Model Context Protocol (MCP) to build applications that connect LLMs to external tools, data sources, and prompt-resources in a standardized way.

What You’ll Learn

• The purpose and architecture of the Model Context Protocol (MCP), how it enables AI apps to integrate with external tools/data without bespoke integrations.
• How to build an MCP-compatible server exposing tools, prompt templates, and resources (e.g., local files, web scraping, data repositories).
• How to build an MCP client inside an AI application (chatbot or agent) that connects to MCP servers and uses their capabilities.
• How to deploy and configure MCP servers and clients, including connecting to reference servers (e.g., filesystem, fetch) and integrating with applications like Claude Desktop.
• Best practices for building rich-context AI applications that scale and avoid fragmented tool integrations.

Setup Instructions

1. Clone the repository

2. Install dependencies

   pip install fastmcp mcp-sdk python-dotnet (or other SDKs as needed)
   

   (Ensure you check the requirements.txt or code comments for precise packages.)

3. Run the examples

   • Open notebooks and run sequentially.

   • To experiment with the server/client patterns:

     python create_mcp_server.py
     python create_mcp_client.py
     

   • Explore connecting to reference servers or external tools via the MCP protocol.

How to Get Started

• Start with the “Introduction to MCP” notebook which covers the background and architecture of MCP.
• Then move to the “Server Setup” notebook/script to build your own MCP server exposing tools/resources.
• Next, open the “Client Setup” notebook/script to connect an AI application to your server.
• Finally, use the “Deployment & Integration” examples to connect multiple servers, configure clients like Claude Desktop, and deploy your system remotely.
• You’re encouraged to customize: add new tools, data sources, or connect to other MCP-compatible servers.

Why This Matters

As AI systems become more ubiquitous, the ability to seamlessly integrate models with tools and data sources is critical. MCP provides a standard protocol for doing so — reducing integration complexity, promoting reuse, and enabling richer AI applications. This course and repo give you the practical skills to build next-generation AI systems that truly connect with the world.