MCP Servers: connect AI apps to tools, then build your own

MCP Servers is an interactive MCP course about the Model Context Protocol, the open standard that lets AI apps use outside tools and data. In 25 short sections it covers why MCP exists, hosts, clients and servers, the three things a server can offer, the JSON-RPC messages underneath, local and remote transports, adding a ready-made server, approvals and safety, and building, testing and sharing your own server in TypeScript or Python.

MCP is how an assistant such as Claude can read your files, query a database or control another app. Once you understand it, you can plug in thousands of existing servers and write one for anything you use.

What the MCP course covers

Why one protocol helps

Without a shared standard, every AI app needs custom code for every tool: three apps and four tools means 3 × 4 = 12 integrations. With MCP, each app speaks MCP once and each tool offers one MCP server, so the same three apps and four tools need only 3 + 4 = 7 pieces.

How it works

The host is the AI app you use, such as Claude Desktop, VS Code or Cursor. It runs a client for each server it connects to. A server can offer tools (actions the model can call), resources (data it can read) and prompts (ready-made templates). They talk in small JSON-RPC 2.0 messages, carried either over stdio, for a server running on your own computer, or Streamable HTTP, for a remote one.

Building your own

The official SDKs make a server a few dozen lines. You register each tool with a name, a description and a schema for its inputs, test it in the MCP Inspector, then share it remotely. The course is firm on safety: a local server runs with your permissions, so install only trusted servers, give each the least access it needs and read approval prompts before allowing them.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": { "name": "get_weather", "arguments": { "city": "Leeds" } }
}

Try a question from the course

Decide what the model does with the feature. If it performs an action, such as sending an email or creating a calendar event, it is a tool. If it is data to read, such as the contents of a file, it is a resource. If it is a reusable template the user picks, such as "summarise this meeting", it is a prompt.

Now classify a few features, then press Mark to check your answers in the course.

Who this course is for

• Developers who want their app, API or data available inside AI assistants.
• Power users who want to connect Claude, Cursor or VS Code to their files, GitHub or databases safely.
• Teams deciding whether to build an MCP server, who need to understand what it involves.

What you need to know first

The first half needs no coding, just comfort with editing a JSON config file. The build sections use TypeScript or Python, so some programming helps; the APIs course covers JSON and requests, and AI Prompt Engineering covers how models use tools.

How long it takes

There are 25 sections and most take a few minutes. Learners without a plan can open two sections of a course a day, in order, so the whole course takes 13 days. A plan removes the daily limit and lets you go straight to building a server.

Read next

AI · Using AIHow to use MCP servers: connect AI apps to your toolsHow to use MCP servers: what MCP is, how hosts, clients and servers fit together, adding a local or remote server to an AI app, and using them safely.