Model Context Protocol (MCP)

Low-Code & Open Protocols

Open Standard for Model Context and Tool Access

Maintained by Anthropic / Open Standard

Core Architecture

MCP is a client-server architecture developed to establish a standard interface between LLM applications (clients) and data sources/development tools (servers). The protocol operates over JSON-RPC 2.0 (transported via SSE or Stdio), allowing servers to expose Resources (static data sources), Prompts (templates), and Tools (executable functions) to the client model dynamically.

How to Use & Configuration

code_example.jsjavascript

import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";

const server = new Server({ name: "weather-mcp", version: "1.0.0" }, {
  capabilities: { tools: {} }
});

server.setRequestHandler(ListToolsRequestSchema, async () => ({
  tools: [{ name: "get_weather", description: "Fetch weather details" }]
}));

const transport = new StdioServerTransport();
await server.connect(transport);

Technology Payment Plans

Specification StandardFree

Open-source standard. Completely free to read, adapt, and build servers or clients.

Community ServersFree

Access the growing list of community-built integrations (databases, search tools, GitHub) without cost.

Key Advantages

•Standardizes how tools and context are exposed to AI models, avoiding custom APIs
•Allows hot-swapping different LLM clients (Cursor, Claude Desktop) with the same tools
•Highly modular, open-source standard with a rapidly growing list of servers

Comparison Analysis

Technology	Primary Use Case & Engineering Focus
Model Context Protocol	Standardized open protocol for context sharing and cross-editor tool compatibility
Ad-hoc Custom APIs	Custom APIs require writing unique schemas and connection code for every client, whereas MCP provides a drop-in bridge.