MCP for Ecommerce Part 2: Build a Real Shopping Agent in 15 Minutes

Part 1 covered why ecommerce needs MCP infrastructure. This part shows you how to build an agent that actually shops.

---

You have an MCP server. You have product data. Now what?

This tutorial walks you through building a real AI shopping agent — one that searches products, compares prices across markets, and finds the best deals. All in 15 minutes.

What You'll Build

A shopping agent that can:

• Search for any product across 50M+ items
• Compare prices across US, Singapore, Japan, Korea markets
• Find active deals and promotions
• Return structured results your UI can render

Prerequisites

• Node.js 18+
• A free BuyWhere API key (get one here)
• An MCP client (Claude Desktop, Cursor, VS Code, or a custom agent)

Step 1: Install the MCP Server (30 seconds)

npx -y @buywhere/mcp-server

This starts the MCP server. It exposes 5 tools to any connected AI agent.

Step 2: Configure Your MCP Client (2 minutes)

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "buywhere": {
      "command": "npx",
      "args": ["-y", "@buywhere/mcp-server"],
      "env": {
        "BUYWHERE_API_KEY": "your-api-key-here"
      }
    }
  }
}

Cursor / VS Code

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "buywhere": {
      "command": "npx",
      "args": ["-y", "@buywhere/mcp-server"],
      "env": {
        "BUYWHERE_API_KEY": "your-api-key-here"
      }
    }
  }
}

Step 3: Your First Query (1 minute)

Ask your agent:

"Find me the best laptop under $1000 available in the US right now."

The agent will call search_products with filters and return real-time results.

Step 4: Cross-Market Price Comparison (3 minutes)

Now the interesting part. Ask:

"Compare the price of Sony WH-1000XM6 headphones across US, Singapore, and Japan. Show me the cheapest option including shipping estimates."

The agent calls compare_products across markets and returns a structured comparison.

Example response structure:

{
  "comparisons": [
    {
      "market": "US",
      "price": "$348.00",
      "retailer": "Amazon",
      "in_stock": true
    },
    {
      "market": "Singapore",
      "price": "SGD 459.00",
      "retailer": "Shopee",
      "in_stock": true
    },
    {
      "market": "Japan",
      "price": "¥42,800",
      "retailer": "Rakuten",
      "in_stock": true
    }
  ]
}

Step 5: Deals & Promotions (2 minutes)

Ask:

"Show me active deals on gaming monitors this week. Sort by discount percentage."

The agent calls get_deals and returns live promotions.

Step 6: Build a Custom Agent (5 minutes)

Here's a minimal Node.js agent that uses the MCP server programmatically:

import { Client } from '@modelcontextprotocol/sdk/client/index.js';
import { StdioClientTransport } from '@modelcontextprotocol/sdk/client/stdio.js';

const transport = new StdioClientTransport({
  command: 'npx',
  args: ['-y', '@buywhere/mcp-server'],
  env: { BUYWHERE_API_KEY: process.env.BUYWHERE_API_KEY }
});

const client = new Client({
  name: 'my-shopping-agent',
  version: '1.0.0'
});

await client.connect(transport);

// Search for products
const result = await client.callTool({
  name: 'search_products',
  arguments: {
    query: 'mechanical keyboard under $100',
    market: 'US',
    limit: 5
  }
});

console.log(result.content[0].text);

The 5 MCP Tools at Your Disposal

Tool	What It Does
search_products	Natural language search with price, category, market filters
get_product	Full product details — specs, price history, availability
compare_products	Cross-market price comparison
get_deals	Active promotions and price drops
get_categories	Category taxonomy and browse structure

Real World Examples

Price tracking bot: Ask your agent to monitor a product daily and alert you on price drops.

Gift finder: "Find a birthday gift under $50 for a gamer in Singapore."

Travel shopper: "Compare electronics prices US vs Japan — I'm traveling next week."

Ecommerce research: "What are the top 5 best-selling air purifiers in Singapore under SGD 300?"

Why This Matters

MCP is becoming the standard for agent-tool communication. Ecommerce is the single largest category of tools humans use online. Connecting them is inevitable — and BuyWhere makes it real today.

Every AI agent should have access to real product data. Not scraped HTML. Not stale datasets. Not crypto-only catalogs. Real, structured, merchant-direct data.

Get Started

npx -y @buywhere/mcp-server

• GitHub: github.com/buywhere/buywhere-mcp
• API Key: buywhere.ai/api-keys
• Docs: docs.buywhere.ai

---

Built with BuyWhere MCP. Star us on GitHub — it helps agents find us.