Skip to content

Model Context Protocol (MCP)

What It Is

An open standard (by Anthropic) for connecting AI assistants to external tools and data sources. Think of it as USB for AI tools — a standard interface so any assistant can use any tool.

Why It Matters

Before MCP, every AI tool had its own integration method. Want Claude Code to access your database? Write a custom tool. Want it to search Slack? Another custom integration. MCP standardizes this:

Without MCP:                    With MCP:
┌────────┐  custom  ┌────┐     ┌────────┐  MCP   ┌────────────┐
│ Claude │─────────→│ DB │     │ Claude │───────→│ MCP Server │
│  Code  │  custom  ├────┤     │  Code  │        │ (any tool) │
│        │─────────→│Slack│    │        │        └────────────┘
│        │  custom  ├────┤     │ Cursor │  MCP   ┌────────────┐
│        │─────────→│FS  │     │        │───────→│ Same server│
└────────┘          └────┘     └────────┘        └────────────┘

Protocol Details

Transport

Two transport modes: - stdio: Server runs as a local process. Communication via stdin/stdout. Best for local tools. - SSE (Server-Sent Events): Server runs as HTTP service. Best for remote/shared tools.

Message Format

JSON-RPC 2.0 over the transport layer:

// Client → Server: tool call
{"jsonrpc": "2.0", "method": "tools/call", "params": {"name": "query_db", "arguments": {"sql": "SELECT..."}}, "id": 1}

// Server → Client: result
{"jsonrpc": "2.0", "result": {"content": [{"type": "text", "text": "..."}]}, "id": 1}

Capabilities

An MCP server exposes: - Tools: Functions the model can call (like API endpoints) - Resources: Data the model can read (like files or database records) - Prompts: Reusable prompt templates

Building an MCP Server

Minimal Python server using the official SDK:

from mcp.server import Server
from mcp.types import Tool, TextContent

server = Server("my-server")

@server.list_tools()
async def list_tools():
    return [Tool(name="greet", description="Say hello", inputSchema={...})]

@server.call_tool()
async def call_tool(name, arguments):
    if name == "greet":
        return [TextContent(type="text", text=f"Hello, {arguments['name']}!")]

Configuration

In Claude Code (~/.claude/settings.json): 

{
  "mcpServers": {
    "my-server": {
      "command": "python",
      "args": ["path/to/server.py"],
      "type": "stdio"
    }
  }
}

Ecosystem