Introduction to Model Context Protocol by Anthropic

What is this course about?
Building integrations between AI models and external services often requires writing extensive custom code. The Model Context Protocol (MCP) standardizes this process, allowing Claude to connect to tools, data sources, and APIs through a consistent interface. This course teaches you to build both MCP servers that expose tools, resources, and prompts, and MCP clients that consume them using the Python SDK.
Get started by exploring MCP's three core primitives—tools, resources, and prompts—and how they integrate with Claude AI. Along the way, learn how to build servers and clients, exploring how MCP simplifies connections to external systems. The course concludes with a hands-on project in which you implement a document management system using MCP.

Instructor
Who teaches this course?
Anthropic is a public benefit corporation dedicated to securing its benefits and mitigating its risks. Anthropic is an AI safety and research company that builds reliable, interpretable, and steerable AI systems.

Objectives
What will I be able to do by the end of this course?
Describe MCP architecture and how it shifts tool definition and execution burden from your server to specialized MCP servers.
Define MCP's transport-agnostic communication system and the message types used between clients and servers.
Outline the complete request-response flow from user queries through MCP clients to external services and back to Claude.
Build MCP servers using the Python SDK with decorators to define tools instead of writing JSON schemas manually.
Implement document management functionality with tools for reading and editing documents using field descriptions and type hints.
Use the built-in MCP Inspector to test and debug your server functionality in a browser-based interface.
Define resources for exposing read-only data, including both direct resources with static URIs and templated resources with parameters.
Implement resource reading functionality in clients with proper MIME type handling for JSON and text content.
Build prompts that provide pre-crafted, high-quality instructions for common workflows like document formatting.
Explain when to use each MCP primitive: tools (model controlled), resources (app controlled), and prompts (user controlled).
Examine practical integration patterns including autocomplete functionality and context injection for AI conversations.

Audience
Who is this course for?
Developers looking to create MCP servers
Engineers who want to integrate Claude with external tools and services without writing extensive integration code
Software developers with an interest in AI model integration
Technology enthusiasts aiming to optimize connection processes between AI services and external data sources

Prerequisites
What do I need to know before taking this course?
Basic Python programming experience
Basic familiarity with APIs
Basic familiarity with async and await patterns