BuyWhere MCP Server — product search and comparison for AI shopping agents

[BuyWhere]

I built a Model Context Protocol server that lets AI shopping agents query a real e-commerce product catalog with price, discount, and availability signals.

Canonical MCP surface:

• Server card: https://buywhere.ai/.well-known/mcp/server-card.json
• Streamable HTTP endpoint: https://mcp.buywhere.ai/mcp
• MCP guide: https://api.buywhere.ai/docs/guides/mcp
• API keys: https://buywhere.ai/api-keys/

As of 2026-04-30, https://mcp.buywhere.ai/health reports catalog.total_products: 5000.

The server currently exposes five tools:

• search_products — keyword search with merchant / price / region filters
• get_product — full product details by BuyWhere ID
• compare_products — side-by-side comparison of 2–10 product candidates
• get_deals — products sorted by discount percentage
• list_categories — top-level catalog taxonomy

All responses follow MCP JSON-RPC 2.0 over HTTP POST. Auth uses bearer tokens.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "search_products",
    "arguments": {
      "q": "iphone 15",
      "domain": "lazada",
      "country_code": "SG",
      "limit": 5
    }
  }
}

A matching live response shape looks like:

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "products": [
      {
        "id": "b8f2...",
        "title": "Apple iPhone 15 128GB Blue",
        "price": 1049.0,
        "currency": "SGD",
        "merchant": "lazada",
        "url": "https://...",
        "in_stock": true
      }
    ]
  }
}

Why I thought this was worth sharing here:

• it is a live MCP implementation, not a mock
• the tool surface is narrowly scoped to a real commerce use case
• the transport and tool schema are public, so other builders can inspect or integrate quickly

If useful, I can share more detail on the auth flow or how we structured the tool surface.