What Is MCP (Model Context Protocol)?
The Model Context Protocol (MCP) is an open standard that defines a universal method for connecting AI assistants to external data sources and tools. Launched in late 2024, it simplifies how models like Claude and ChatGPT interact with enterprise systems.
The Four Core Primitives
Tools: Executable functions that models invoke (e.g., querying a database).
Resources: Read-only data sources (e.g., file contents or API responses).
Prompts: Reusable templates that standardize request patterns.
Servers: Lightweight processes that expose these capabilities to clients.
Why It Matters
Before MCP, every integration was bespoke. If you wanted an AI to read your Slack and GitHub, you needed two separate, non-standard connectors. MCP provides a protocol-level abstraction that decouples AI clients from external tool implementations.
Implementation Example (TS)
import { Server } from "@modelcontextprotocol/sdk/server/index.js";
const server = new Server({
name: "weather-server",
version: "1.0.0"
}, { capabilities: { tools: {} } });
// Handler for tools/list...FAQ
Is MCP limited to Anthropic models?
No. It is an open protocol that any client or model can implement.
How does it differ from function calling?
Function calling is model-specific features; MCP is a cross-platform standard for discovery and transport.