Skip to main content

MCP Guides

Practical, tested guides to the Model Context Protocol (MCP): using MCP servers in the AI apps you already have, building and publishing your own, and keeping both secure. Every guide leads with the answer, uses complete code we run before publishing, and shows the versions it was last checked against.

Start here​

I want to use MCP servers​

  1. Learn what MCP is and how it works
  2. Understand what MCP adds over a plain API
  3. Pick an AI app that supports MCP
  4. Find servers for the work you do
  5. Check a server before you install it
  6. Connect your first server (Claude Desktop shown)

I want to build an MCP server​

  1. Decide whether to write, generate or use a hosted server
  2. Build your first MCP server
  3. Write tools the model picks correctly
  4. Test it with MCP Inspector
  5. Add automated tests and CI
  6. Deploy it as a remote server
  7. Run it in production
  8. Publish it to the MCP Registry

All sections​

Use MCP servers

Connect servers to your AI app, sign in with OAuth, and fix connections.

Discover servers

Where to find good servers, official servers by use case, and how the registry works.

Build & publish

From a first server to a published, authorized, remote one.

Tool design

Tools the model picks correctly and calls successfully.

Security

The attacks that actually happen and the defences that work.

MCP Apps

Interactive interfaces that render inside AI apps.

WebMCP

Let browser agents use your site through tools you define.

Read how we test these guides to see what each example checks and how often.