What is MCP (Model Context Protocol)?

How MCP works

MCP has three roles. The host is the AI application the user works in (Claude Desktop, an IDE, your own app). Inside it, an MCP client connects to one or more MCP servers, each of which wraps a system such as a database, a code repository or your product’s API.

A server offers three kinds of things:

• Tools: actions the model can call, like search_orders or create_ticket, each with a JSON Schema for its inputs.
• Resources: read-only data the client can load into context, like a file or a record.
• Prompts: reusable prompt templates the user can pick.

Messages use JSON-RPC 2.0, over stdio for servers running on the same machine or streamable HTTP for remote servers.

An example

A Laravel app can expose an MCP server with a find_customer tool. Once it is added to Claude or Cursor, you can ask “which customers in Ohio renewed last month?” and the model calls the tool, reads the result and answers, without you writing a chatbot integration.

When to use it

Use MCP when the same capability should be available to several AI clients, or when you want users to plug your product into the AI tools they already use. For a single, fixed integration inside your own app, plain tool calling against your API is often simpler.

Security

An MCP server acts with the permissions you give it. Grant the least access needed, prefer read-only tools, require confirmation for actions that change data, and keep tokens out of shared config files. The MCP Server Config Generator writes client configs with secrets kept out.