Model Context Protocol (MCP) Explained for Developers

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to external tools, data sources and prompts through one common interface. Introduced by Anthropic in November 2024 and adopted broadly across the industry since, it plays the role people kept reaching for with the "USB-C of AI" analogy: implement the port once, and every compatible device just works.

The problem MCP solves

Before MCP, every AI application integrated every tool bilaterally. A coding assistant that needed GitHub, Postgres and Slack shipped three bespoke integrations; a second assistant needing the same tools shipped three more. That is the classic M × N integration explosion.

MCP collapses it to M + N. Tool providers implement an MCP server once. AI applications implement an MCP client once. Any client can then talk to any server.

The three primitives

An MCP server can expose three kinds of capabilities:

• Tools — functions the model can call, with JSON Schema parameters (search_issues, run_query, send_message).
• Resources — readable data identified by URIs that the client can load into context (a file, a database row, a dashboard).
• Prompts — reusable, parameterized prompt templates the server offers to the client.

Under the hood the protocol is JSON-RPC 2.0, carried over stdio for local servers or streamable HTTP for remote ones. (The transport layer is evolving fast — the 2026 spec revision makes the protocol core stateless, which simplifies running remote servers at scale.) A minimal TypeScript server looks like this:

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { z } from "zod";
 
const server = new McpServer({ name: "weather", version: "1.0.0" });
 
server.tool(
  "get_forecast",
  { city: z.string() },
  async ({ city }) => ({
    content: [{ type: "text", text: await fetchForecast(city) }],
  }),
);

Why adoption happened so fast

Three things lined up. First, the spec shipped with working SDKs and dozens of reference servers, so there was something to run on day one. Second, it was genuinely open — when OpenAI and Google DeepMind added client support in 2025, MCP stopped being a vendor feature and became infrastructure. Third, the timing matched the shift from chatbots to agents: an agent is only as useful as the systems it can touch, and MCP made "touching systems" a solved problem.

What to watch out for

MCP moves the integration problem; it does not remove the security problem. A server runs with real credentials against real systems, so the usual rules apply: least-privilege tokens, careful review of third-party servers before connecting them, and human confirmation in front of destructive operations. Prompt injection through tool results remains an active research area — treat data returned by tools as untrusted input, not as instructions.

For developers, the practical takeaway is simple: if you maintain a product that AI assistants should be able to use, shipping an MCP server is now the default way to make that happen.