Model Context Protocol: Advanced Topics
Anthropic
Last updated October 11, 2026
Model Context Protocol: Advanced Topics is a free course from Anthropic that picks up where the introductory course leaves off and covers the technical implementation of Model Context Protocol (MCP) servers and clients — from message passing to production deployment. It works through the features that make a server genuinely capable: sampling, which lets a server ask the client's model to run a request (and so shifts that cost to the client); progress notifications for long-running work; and the roots permission model for handling file access safely. From there it digs into how messages actually flow — the JSON-RPC message types, the stdio transport for local development, and the streamable HTTP transport for remote and scalable deployments, including when you trade features for scale. It closes on scaling: when to run a server statelessly over HTTP so it can sit behind a load balancer, and what you give up compared with a stateful one.
What you'll learn
- Implementing MCP servers with tool functions, logging, and progress notifications
- Handling bidirectional communication between MCP clients and servers
- Configuring file-system access safely with the roots permission model
- Using sampling callbacks so a server can initiate model requests through the client
- Working with both stdio and HTTP transports for local and remote deployments
- Telling request-and-result messages apart from notifications in MCP's message format
- Deciding when to run a stateless HTTP server so it can scale behind a load balancer
Frequently asked questions about Model Context Protocol: Advanced Topics
Who is Model Context Protocol: Advanced Topics for?
Engineers building production MCP servers who need to understand the protocol's advanced capabilities.
Is Model Context Protocol: Advanced Topics free?
Yes — Model Context Protocol: Advanced Topics is completely free to take.
What are the prerequisites for Model Context Protocol: Advanced Topics?
A basic understanding of MCP servers and clients, and familiarity with async programming patterns — the introductory MCP course, or equivalent experience, is the natural lead-in.
Do you need to code for Model Context Protocol: Advanced Topics?
Yes — Model Context Protocol: Advanced Topics involves hands-on coding.
Why we suggest this course
The course for developers past the basics of MCP who need to ship a server that holds up in production. It concentrates on exactly the things the introduction leaves out — sampling, notifications, the roots permission model, the difference between stdio and streamable HTTP transports, and the trade-offs involved in going from a local prototype to a scalable service. Two things worth knowing: it assumes you already understand MCP servers, clients, and async patterns — Introduction to Model Context Protocol is the lead-in if the basics are new; and it's taught in Python with Anthropic's MCP tooling and Claude as the client, though the servers you build run independently of any one vendor.