Tech World Software SchoolTWSS
0x00Lesson 1 of 16

Understand the Model Context Protocol

See the problem MCP solves and the roles of hosts, clients and servers.

15 min 6-question quiz 1 code exercise
By the end of this lesson you can
  • Explain the integration problem MCP solves
  • Describe the roles of an MCP host, client and server
  • Tell local servers from remote ones

An AI assistant becomes far more useful when it can read your files, check your calendar or open a ticket. Before MCP, every AI application wrote its own integration for every service: 4 apps × 5 services = 20 custom connectors, each built and maintained separately.

The Model Context Protocol (MCP) is an open standard for that connection - often compared to a USB-C port for AI applications. A service implements MCP once and works with every MCP-capable app; an app implements MCP once and can use every MCP server. 4 + 5 = 9 implementations instead of 20.

integrations.py
apps, services = 4, 5
print("custom connectors:", apps * services)
print("with a shared protocol:", apps + services)
Output
custom connectors: 20
with a shared protocol: 9

Hosts, clients and servers

  • The host is the AI application you use - a chat app, an IDE, an agent. It owns the user experience, talks to the language model, and decides what the model may do.
  • For each server it connects to, the host runs an MCP client that speaks the protocol to that one server.
  • An MCP server exposes capabilities: tools to call, resources to read, prompts to use.

A server can run locally - launched by the host as a subprocess, talking over standard input/output - or remotely, reached over HTTP. MCP standardizes the conversation; it does not decide which model the host uses, and connecting to a server does not make that server trustworthy.

Try it

Up to 15 XP

Who does what?

Sort each responsibility to the part of an MCP setup that owns it.

0 of 6 sortedScore 0/0

  • “Ask the user to confirm before a tool sends an email”

  • “Send a tools/call request over the connection to one server”

  • “Expose a “create_ticket” tool and actually create the ticket”

  • “Choose which language model to use”

  • “Validate the arguments of a tool call before running it”

  • “Combine the tool lists of three servers for the model”

Key takeaways

  • MCP turns N × M custom integrations into N + M implementations of one protocol.

  • The host is the app and decision-maker; it runs one client per server; servers expose capabilities.

  • Servers run locally (subprocess) or remotely (HTTP); speaking MCP doesn’t make a server trustworthy.

Lesson quiz

6 questions · pass with 5 correct · up to 50 XP

Passing this quiz completes the lesson and keeps your streak going. Questions you miss come back in review sessions later.

Practice: build an MCP server

Write Python in the editor and run it against sample inputs. Python runs locally in your browser using a WebAssembly runtime.

Exercise 1

Count the integrations

+25 XP

Read the number of AI applications and the number of services on one line. Print custom: X - one connector per app and service pair - and then mcp: Y - one MCP implementation per app and per service.

  • 4 apps, 5 services
  • 1 app, 1 service
main.py
Loading editor…

Python runs in a sandboxed browser worker with a 60 second time limit. Its runtime loads from the Pyodide CDN; your code stays in this browser.

Questions about this lesson

Stuck? Ask. Figured something out? Share it. Explaining is one of the best ways to learn.

Loading posts…