MCP — Connecting External Tools to AI Coding Agents

Model Context Protocol (MCP) is an open standard, originally published by Anthropic in late 2024, that lets AI agents talk to external tools using a common JSON-RPC interface. By April 2026 it has hit roughly 100 million monthly downloads with 3,000+ indexed servers and is supported by Claude Code, Codex CLI, Cursor, and Windsurf.

Why this matters

Before MCP, every AI tool invented its own plugin system. If you wrote a GitHub integration for one IDE, it didn't work anywhere else. MCP is the common adapter — one server speaks to every compliant client. The USB-C analogy is apt.

The three primitives

• Tools: functions the agent can call (query_db, create_issue, send_slack_message)
• Resources: read-only data sources the agent can pull from (files, docs, schemas)
• Prompts: reusable prompt templates the server exposes as commands

A minimal MCP server in TypeScript

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import { z } from "zod";

const server = new McpServer({
  name: "weather",
  version: "1.0.0",
});

server.tool(
  "get_forecast",
  {
    city: z.string().describe("City name"),
  },
  async ({ city }) => {
    const res = await fetch(`https://api.weather.example/${city}`);
    const data = await res.json();
    return {
      content: [{ type: "text", text: JSON.stringify(data) }],
    };
  },
);

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

Wiring a server to a client

{
  "mcpServers": {
    "weather": {
      "command": "node",
      "args": ["/path/to/weather-server/index.js"]
    },
    "postgres": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres", "postgresql://localhost/mydb"]
    }
  }
}

Servers worth knowing in 2026

Security model

• Servers run locally by default — no mystery cloud calls
• The client mediates every tool invocation — user can approve or deny
• Auth flows through the server (API keys, OAuth) — not exposed to the model
• Audit logs of tool calls are the minimum observability you should require

“Protocols are patient. They don't win today; they win in five years.”

The big idea: MCP turned AI coding tools from siloed islands into a composable ecosystem. If you build, extend, or integrate AI agents in 2026, this is the protocol layer you cannot ignore.