Model Context Protocol (MCP)

The Model Context Protocol (MCP) is an open standard introduced by Anthropic in late 2024. It solves a simple but annoying problem: how to connect an AI assistant to the tools and data it needs - a database, a file system, an API, a calendar, an internal company system. Without a shared standard, every developer had to write a custom integration for each combination of model and tool. MCP defines a single interface, so a tool written once works with any client that supports the protocol.

A common analogy: MCP is to AI applications what USB-C is to hardware. Instead of dozens of incompatible connectors, one standard. In MCP terminology, a server exposes capabilities (tools the model can call, resources it can read, and prebuilt prompts), and a client is the AI application that uses them.

Technically MCP is built on JSON-RPC. A server can run locally or remotely and communicates over standard input and output or over HTTP. That makes the protocol a building block for so-called agentic applications - AI systems that don't just answer questions but take steps themselves: read a document, query a database, create a record, send an email.

For businesses, MCP matters mainly because it lowers the cost of integrating AI into existing systems. Once your internal tool has an MCP server, you can build an assistant on top of it without rewriting the underlying logic. The standard was adopted quickly by other companies and has become a de facto common language for connecting models to the world around them.