MCP Architecture Expert Skill

Purpose

Master the Model Context Protocol (MCP) to build standardized, reusable integrations between AI systems and data sources, eliminating the N×M integration problem.

What is MCP?

Model Context Protocol

Open standard (November 2024, Anthropic) for connecting AI systems to external data sources and tools through a unified protocol.

The Problem: N agents × M tools = N×M custom integrations The Solution: N agents + M MCP servers = N+M integrations (any agent uses any tool)

Architecture

┌─────────────┐
│  MCP Host   │  (Claude Desktop, IDEs, Apps)
│   ┌─────┐   │
│   │Client│──┼──┐
│   └─────┘   │  │
└─────────────┘  │
                 │ JSON-RPC 2.0
                 │
┌────────────────┼─────────────┐
│  MCP Server    ▼             │
│  ┌──────────────────┐        │
│  │  Resources       │        │
│  │  Tools           │        │
│  │  Prompts         │        │
│  └──────────────────┘        │
│         │                    │
│         ▼                    │
│  ┌──────────────────┐        │
│  │ Data Source      │        │
│  │ (DB, API, Files) │        │
│  └──────────────────┘        │
└─────────────────────────────┘

Three Core Capabilities

1. Resources

Purpose: Expose data for AI to read

Examples:

• File contents
• Database records
• API responses
• Documentation

Definition:

{
  "resources": [
    {
      "uri": "file:///docs/api-spec.md",
      "name": "API Specification",
      "mimeType": "text/markdown"
    },
    {
      "uri": "db://customers/12345",
      "name": "Customer Record",
      "mimeType": "application/json"
    }
  ]
}

2. Tools

Purpose: Functions AI can invoke

Examples:

• Query database
• Call external API
• Process files
• Execute commands

Definition:

{
  "tools": [
    {
      "name": "query_database",
      "description": "Execute SQL query on customer database",
      "inputSchema": {
        "type": "object",
  ...