What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard, introduced by Anthropic in late 2024 and now adopted across the industry, for connecting AI applications to the tools, data, and systems they need. Instead of building a bespoke integration for every model and every service, you build one MCP server per service and any MCP-compatible application can use it.

The one-sentence version #

MCP is a universal connector for AI. It defines how an application (the client) asks an external system (the server) what it can do, and how it invokes those capabilities. Think of it as a common port that any tool can plug into, the way USB-C is a common port for hardware.

The problem it solves #

Before MCP, connecting a model to a service meant custom glue code, unique to that model's tool-calling format and that service's API. Every new model or service multiplied the integration work. MCP replaces that N-times-M problem with N-plus-M: write one server for a service, and every client speaks to it the same way.

For a fuller motivation, see Introducing MCP.

What it is not #

• It is not a model or an API from any single vendor. It is a protocol, like HTTP.
• It is not tied to one language. There are SDKs in Python, TypeScript, and more.
• It is not only for Claude. Many AI applications and IDEs act as MCP clients.