The Model Context Protocol in plain terms

The Model Context Protocol, or MCP, is a standard way for an AI application to discover and use tools and data that live somewhere else.

The problem it solves

Without a standard, every application writes its own glue code for every system it wants to reach. Ten applications and ten systems means up to a hundred integrations. With a shared protocol, each system exposes itself once and each application learns one way to talk to it.

The pieces

• Server. A small program that exposes capabilities. It might wrap a database, an internal API or a set of files.
• Client. Lives inside the AI application. It connects to servers and passes what they offer to the model.
• Tools. Actions the model can call. Each has a name, a description and a schema for its inputs.
• Resources. Data the application can read, such as a document or a record.

How a call flows

1. The client connects to a server and asks what tools it has.
2. The tool names, descriptions and schemas are given to the model.
3. The model decides to call a tool and produces arguments.
4. The client sends the call to the server and returns the result to the model.

Servers can run as a local process or be reached over HTTP, depending on the setup.

Why it matters

The model does not need custom code for each system. It reads the tool descriptions and uses them. That also means the quality of those descriptions decides how well the model uses your tools, which is the subject of the next post.