What Is MCP? A Five-Minute Read

Two Analogies

The Restaurant

Imagine yourself as a diner sitting at a restaurant. You already know that a waiter will come to your table, take your order and bring it to you. That’s an expected behavior, a protocol.

But there’s a problem: you don’t know what dishes the restaurant offers. The waiter can take your order, but that is of little use if you don’t know what you can ask for. So you need a menu.

The menu tells you what is available. You choose what you want, and the waiter takes that request to the restaurant and brings it to you.

Now let’s translate the analogy:

• You, the AI, already know the protocol: you know what to expect.
• You are at the restaurant, the platform, and you use the menu, the MCP, to discover what you can ask the waiter, the API, to do.
• The waiter is the API. The menu is the MCP.

That’s it!

You know what to expect from the waiter. You know what the restaurant can provide because you have the menu. And now you can actually place your order. That’s the basic idea behind MCP.

The Employee

Laura has just joined a company.
On her first day, she is given access to the company’s business system. She has the necessary permissions. She can log in, read information, and perform the actions she is authorized to perform.

But there is a problem.

Having access doesn’t tell Laura what she can do, or how to do it.
She needs to know what is available to her.

Can she add a customer?
Can she update a project?
Can she schedule a meeting?

Now let’s translate:

• Laura is the AI.
• The company’s system is the platform.
• Her access gives her the ability to interact with the system. An API is one common way software can access those capabilities from outside the platform.
• And the information that tells Laura which tools are available and what she can do with them is the role played by MCP.

Laura doesn’t need to learn how the company’s entire technology infrastructure works. She just needs to know what tools are available to her and how to use them.

That’s essentially what MCP helps an AI do.

Bringing the Analogies Together

Now let’s put the three ideas side by side.

Restaurant:
You → Menu → Waiter → Restaurant

Employee:
Laura → Instructions → Her work → Company’s system

Artificial Inteligence:
AI → MCP → Available tools → Platform or system

The idea is surprisingly simple:

1. You need the menu to know what you can ask the waiter.
2. Laura needs instructions to know what she can do.
3. AI uses MCP to understand what it can ask the available tools to do.

The technology behind it can become considerably more complicated, but the basic idea doesn’t have to be.

So, What Is MCP Really?

Now that we’ve seen the idea from two different perspectives, we can finally look at the technical definition.

The official MCP specification defines it this way:

“Model Context Protocol (MCP) is an open protocol that enables seamless integration between LLM applications and external data sources and tools.”