Model Context Protocol (MCP)

Every tool an agent uses has to be plugged in somehow, and historically each app, model, and tool spoke its own private dialect — so connecting N tools to M apps meant N times M custom adapters. The Model Context Protocol is an open standard that fixes this with a common language: a tool exposes itself once, in the agreed format, and any compatible LLM application can discover and use it without bespoke glue. Think of it as a universal port that any tool and any client can plug into.

Technically it follows a client-server shape. A tool or data source runs an MCP server that advertises what it offers — available tools, resources to read, and reusable prompts. The LLM application is the client; it connects, sees the catalogue, and lets the model call those capabilities through one consistent interface. Swapping in a new tool no longer means rewriting how the model talks to it.

The point is interoperability, not new model abilities — MCP standardises the wiring so an ecosystem of tools and apps can mix and match. As with any connector that grants real access, what a server exposes is also an attack surface, so connections are meant to be scoped and trusted.