Lesson 162 · 3 min
Inside MCP: servers, tools and resources
An MCP server is a small program that describes what it can do, and the assistant reads that description to decide which tool to call.
MCP has two sides. The client is the AI app you talk to. The server is a small program, run by an app maker or by you, that sits in front of some app or data and speaks the MCP language.
When they connect, the server lists what it offers. Tools are actions with a name, a plain-language description and the inputs they need, like "create_event: needs a title, a date and a time". Resources are things to read, such as files or records. Some servers also share ready-made prompts.
The model reads those descriptions the same way it reads your message. When you ask for a meeting next Tuesday, it picks create_event, fills in the inputs, and the client sends the call to the server. The server does the real work and returns the result.
This shows why descriptions matter. A vague tool description leads to wrong calls, and a dishonest one can mislead the model. Builders should write tool descriptions as carefully as instructions for a person.
Do it now
Write a tool menu
- 1. Pick an app you use, like a to-do list.
- 2. Write three tools it could offer: a name, one sentence of description, and the inputs needed.
- 3. Ask a chatbot to play the model and choose a tool for a request, using the prompt below.
Pretend you are an AI that can call these tools: [PASTE YOUR THREE TOOLS] For the request "[A REQUEST]", say which tool you would call and with which inputs. Then tell me how my descriptions could be clearer.
Quick check
Words from this lesson
- MCP
- An open standard that lets AI assistants plug into apps and data sources, like a universal adapter for tools.
- Tool use
- An AI's ability to call outside tools, such as a calculator, web search or calendar, to do things it cannot do alone.
- API
- A doorway that lets one program send requests to another, for example an app asking an AI model for an answer.
- Context window
- The maximum amount of text an AI can look at in one go. Anything beyond it is forgotten for that conversation.