District 13 · Plug Station

🏙️ Journey

0/13

🔌

⚡

🔧

🌐

District 13 · Plug Station

Welcome to Plug Station

MCP is the universal plug that lets any AI connect to any tool — no custom wiring required.

Plug In

📁

Resources

🔧

Tools

💬

Prompts

The Integration Problem

Before MCP, every AI tool needed its own custom integration. With MCP, one protocol connects everything.

Each AI client needs custom integration code for each tool:

Claude + GitHub custom

Cursor + DB custom

GPT + Files custom

Claude + DB custom

GPT + GitHub custom

Cursor + Files custom

6 custom integrations for 3 clients × 3 tools. Add one more tool: +3 more integrations 😱

MCP Primitives

MCP servers expose exactly three building blocks — simple, but powerful enough for anything.

📁

Resources

Expose data that the AI can read — files, database rows, API responses, calendar events.

resource: file://notes/2024-01-01.md → returns the file content

Build Your First MCP Server

Hover the highlighted tokens to understand each part of a minimal MCP server.

from mcp.server.fastmcp import FastMCP

# Initialize the server

mcp = FastMCP("weather-server")

@mcp.tool()

def get_weather(city: str) -> str:

"""Get the weather for a city."""

return fetch_weather_api(city)

mcp.run()

← Hover a highlighted token to learn what it does.

The MCP Ecosystem

MCP is open source and growing fast — plug any server into any client.

🟣Claude Desktop

Official

Anthropic's official desktop app has MCP built-in — just add a server config and Claude can access your files, calendar, git repos, and more.

⚡Cursor

Popular

The AI code editor supports MCP servers for repository-aware coding — give Cursor access to your CI/CD, issue tracker, and database schema.

🤖Your Custom Agent

You!

Any Python or TypeScript agent using the MCP SDK can connect to any MCP server — build internal tools your whole team uses.

🌐1,000+ Servers

Community

The open-source MCP registry has servers for GitHub, Slack, Notion, Linear, Postgres, browser automation, and hundreds more ready to plug in.

Mission: Route the Requests

Match each AI request to the correct MCP server that should handle it.

🤖 “Read my local notes.txt file”

🤖 “Open a pull request with my changes”

🤖 “Scrape prices from this product page”

🤖 “Run SELECT * FROM users WHERE active=1”

Get started

Build your first MCP server 🔌

Four steps to connect your tools to any AI client.

1
Install the MCP SDK
Available for Python and TypeScript. Use FastMCP for the simplest DX.
```
pip install mcp fastmcp
# or for Node.js:
npm install @modelcontextprotocol/sdk
```

Define your server

Create a FastMCP server and expose your data sources or actions.

from mcp.server.fastmcp import FastMCP

mcp = FastMCP("my-tools")

@mcp.resource("memo://{id}")
def get_memo(id: str) -> str:
    return db.get_memo(id)

@mcp.tool()
def create_memo(title: str, body: str) -> str:
    db.create(title, body)
    return "Created!"

Expose a tool

Tools are callable actions. Resources are readable data. Both are auto-discovered by clients.

# Start the server (stdio for local, HTTP for remote)
mcp.run()  # stdio mode
# or
mcp.run(transport="sse", host="0.0.0.0", port=8080)

Connect Claude Desktop

Add your server to Claude's config file and restart. Claude can now use your tools.

// ~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "my-tools": {
      "command": "python",
      "args": ["path/to/server.py"]
    }
  }
}

Ask Popy

Chat with the Connector ✨

Questions about MCP servers, tools, resources, or integrations?

Hi! I'm Popy 🔌 Ask me anything about MCP — the protocol, building servers, connecting clients, or the ecosystem!

🎉 Journey Complete!

You've explored all 13 districts of AI City — from writing your first prompt to building production-ready AI systems with evals, observability, guardrails, and standardized tool connections. The city is yours to build!

🏙️ Return to AI City 🎫 View My Passport

Mini Project

Build Quest