~/ § TUTORIAL Model Context Protocol

What is MCP? — Model Context Protocol explained for builders

· 6 min read · English

MCP is the open protocol from Anthropic that lets AI agents like Claude talk to external tools — databases, n8n, Google Sheets, your internal systems — through one common language, instead of writing a new integration every time.

What MCP actually is

MCP stands for Model Context Protocol. It's an open standard, released by Anthropic in late 2024, that lets AI agents connect to external tools in a uniform way.

Analogy: before USB-C, every laptop maker shipped its own port and you needed an adapter for every peripheral. Once USB-C became the standard, everything plugged into one shape. MCP is "USB-C for AI" — every tool speaks the same language, every AI that supports MCP can call them right away. No more writing a fresh adapter per integration.

Why MCP matters for AI automation

Picture the old world: you want Claude to read from Google Sheets, query Postgres, and trigger an n8n workflow. That's three custom integrations. Switch to a different LLM later, and you write all three again.

MCP fixes that: write the MCP server once, use it from any AI that speaks MCP. Claude Desktop, Claude Code, Cursor — and a growing list of clients — all support it.

What this means for automation builders: a team running n8n + AI can build one MCP server that exposes their company workflows, and every AI client they use gets access immediately. No waiting for Anthropic, OpenAI, or anyone else to ship the integration.

MCP server vs MCP client — what's the difference

  • MCP server = a program that "exposes services" for an AI to call — things like "search database X", "send an email", "trigger n8n workflow Y". It exposes tools and resources to the AI.
  • MCP client = a program the AI runs inside (Claude Desktop, Claude Code, Cursor). It's the conduit between the AI and any MCP server.
  • When the AI agent decides to use a tool, it sends a request through the client → the client speaks MCP (JSON-RPC over stdio or HTTP) to the server → the server responds → the AI uses the result in its reply.

Real-world examples

  • Customer support team: An MCP server connects Claude to the company's knowledge base — the AI answers customers with the latest information, no model retraining required.
  • Content + automation team: An MCP server bridges Claude Code to n8n — let the AI trigger a workflow that generates videos and posts them to YouTube on a schedule.
  • Founder / operator: An MCP server pulls from Stripe + Google Calendar + your CRM — ask Claude "summarize this month's revenue and what's on the calendar" and it pulls live data to answer.

How to start with MCP — for beginners

The fastest path is using existing MCP servers that other people have written, before building your own:

  1. Install Claude Desktop (the friendliest MCP client today) or Claude Code (the CLI).
  2. Browse the catalog at github.com/modelcontextprotocol/servers — ready-to-use servers for filesystem, github, postgres, slack, brave-search, and more.
  3. Edit your config (for Claude Desktop, that's claude_desktop_config.json):
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/path/to/folder"
      ]
    }
  }
}
  1. Restart Claude Desktop → the AI sees the new tools immediately. Try "read the files in that folder and summarize them for me."
  2. When you're ready to write your own MCP server — use the TypeScript SDK @modelcontextprotocol/sdk or the Python SDK. Both have examples in the same repo.

Why it's worth learning MCP now

MCP isn't a short-term trend — it's a standard that will stick around because it solves the integration fragmentation everyone hits. Whoever ships MCP servers now has tools ready before the wave; they're building blocks for the more complex agentic workflows landing this year.

At BoomBigNose+ we build MCP servers from zero, wired into n8n and Claude Code — from the first server that talks to Google Sheets through systems where the AI orchestrates company workflows on its own. Thai-first, with English support.