9.5 Model Context Protocol and interoperability
You can expose your own system as tools any agent can use.
Before:08. Large Language Models
The Model Context Protocol standardises how agents reach external systems: servers expose tools, resources and prompts; clients consume them over defined transports. Writing a server for a real API is the fastest way to understand the contract. It sits here because interoperability is what lets agent work compound. The trust boundary is the security lesson — tool descriptions and results from a server are untrusted input to your agent, however official the server looks.
Work through these
MCP servers, clients, transports
An open protocol for connecting agents to tools and data, with a server offering capabilities and a client consuming them over a defined transport. It exists so every tool need not be integrated separately.
Model Context Protocol — Introduction · DocsTools, resources and prompts as primitives
The three things a server can offer: actions the agent may take, data it may read, and prompt templates it may use. Knowing the distinction shapes how you design a server.
Writing an MCP server for a real API
Writing a server that exposes a real service, which is the exercise that makes the protocol concrete. It is also a genuinely useful portfolio piece.
Model Context Protocol — Introduction · DocsTrust boundaries between agent and server
The agent and the server do not trust each other automatically, and content returned by a server can carry instructions. Treating server output as data rather than commands is the security position.
Sign in to keep your progress.
Free resources
We haven't checked most of these for screen reader use yet.
Links last checked 29 Aug 2026.
Stuck here?
Ask a mentor. A real person answers, and they can see exactly which topic you're on. Usually within a couple of working days.
Checking your session…
Topics shown in module order.