MCP Simulator

The Model Context Protocol (MCP) is how an AI model talks to outside tools — a database, your files, GitHub. It’s hard to picture from a definition, so watch it happen: pick a mock server, hit run, and follow the model through the four steps of a tool call.

A read-only database the model can query with SQL.

Tools this server exposes

queryRun a read-only SQL query and return the rows.
You ask: Find users with active subscriptions
1Model parses the request
Pick a tool

The model reads the server’s tool list, sees a `query` tool that runs SQL, and decides to use it. It writes the SQL itself from your plain-English request.

2Model sends a tools/call
Send the request

The model sends a JSON-RPC tools/call over MCP — naming the tool and its arguments.

3Server returns structured data
Get structured data back

The MCP server runs the query and returns the result as content the model can read — not a screenshot, real data.

4Model writes the final answer
Answer in plain English

The model turns the rows into a normal reply: “There are 3 users with active subscriptions — Ada Lovelace, Alan Turing, and Grace Hopper.”

This is a visual simulation of the Model Context Protocol handshake — the JSON is real MCP, the data is mocked. No live model or server is contacted.

Want the full explanation? Read What Is MCP? for the plain-English version, or browse our MCP explainers and server roundups.