Model Context Protocol (MCP)

The Model Context Protocol is a specification, open-sourced by Anthropic on 25 November 2024 and since adopted well beyond it, for wiring language-model applications to outside capabilities. An MCP client (living inside a host application like an IDE or chat app) talks to one or more MCP servers, each of which exposes some mix of tools (actions the model can invoke), resources (data the model can read), and prompts (reusable templates). The protocol defines the message format, the capability handshake, and — in later revisions — an authorization model.

Why it matters

Before MCP, every AI-to-tool integration was bespoke. MCP made integrations portable, which is why thousands of servers now exist and why a single vulnerable pattern in the protocol or a popular server implementation can affect many deployments at once. The security-relevant consequence is that untrusted content now flows into models through a standard, widely-implemented channel — and the same standardization that helped adoption also standardizes the attack surface.