Q23: Model Context Protocol (MCP): Client-Host-Server Architecture & Dynamic Tool Discovery - AI Engineering & LLMs (50 Qs)

1. What is the Model Context Protocol (MCP)?

Created by Anthropic and established as an open standard, the Model Context Protocol (MCP) provides a universal, open standard for connecting AI assistants and agents to external data sources, business tools, and developer environments. Prior to MCP, every LLM provider and tool developer wrote custom, proprietary plugins (OpenAI Plugins, LangChain custom tools, Claude artifacts).

arduino

┌────────────────────────────────────────────────────────────────────────┐
│                        MCP Architecture                                │
└────────────────────────────────────────────────────────────────────────┘
  Host Application (Claude Desktop, IDE, Agent Framework)
        │
        ├── Client (Manages protocol lifecycle, auth, & tool routing)
        │      │
        │      ├─► Transport (stdio / SSE / WebSocket)
        │      │
        ▼      ▼
  MCP Servers (Isolated Processes / Microservices)
  ┌───────────────────┐    ┌───────────────────┐    ┌───────────────────┐
  │ GitHub MCP Server │    │ Postgres Server   │    │ FileSystem Server │
  │ Tools: PRs, Issues│    │ Tools: Run Query  │    │ Tools: Read, Edit │
  └───────────────────┘    └───────────────────┘    └───────────────────┘

2. Core Primitives of the Protocol

Primitive	Initiator	Purpose	Example
Tools	Model-controlled	Executable functions that cause external side-effects or fetch data.	`query_sales_db(sql)`, `create_jira_issue()`
Resources	Application-controlled	File-like read-only data that provide context to the model.	`postgres://db/schema`, `file:///docs/spec.md`
Prompts	User-controlled	Pre-defined, interactive prompt templates surfaced in the UI.	`debug_incident_template`, `code_review_prompt`

3. Writing an Enterprise MCP Tool in TypeScript

typescript

import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
import { CallToolRequestSchema, ListToolsRequestSchema } from '@modelcontextprotocol/sdk/types.js';

const server = new Server({ name: 'enterprise-ops', version: '1.0.0' }, { capabilities: { tools: {} } });

// Register available tools
server.setRequestHandler(ListToolsRequestSchema, async () => ({
  tools: [
    {
      name: 'get_service_health',
      description: 'Checks Kubernetes cluster health and active incident count',
      inputSchema: {
        type: 'object',
        properties: { serviceName: { type: 'string' } },
        required: ['serviceName']
      }
    }
  ]
}));

// Handle tool execution
server.setRequestHandler(CallToolRequestSchema, async (request) => {
  if (request.params.name === 'get_service_health') {
    const { serviceName } = request.params.arguments as { serviceName: string };
    return { content: [{ type: 'text', text: JSON.stringify({ service: serviceName, status: 'healthy', replicas: 5 }) }] };
  }
  throw new Error('Tool not found');
});

const transport = new StdioServerTransport();
await server.connect(transport);