show_hn

Show HN: BuyWhere — the product catalog API that AI agents actually want to use

Every AI shopping assistant hits the same wall: product data is stuck behind merchant paywalls, captive HTML, and CAPTCHAs. You can scrape for a demo. You cannot scrape for production.

After watching too many developers burn weeks on Shopee and Lazada parsers — only to watch them break on every layout change — we built BuyWhere. It is a structured product catalog API that gives AI agents real prices, real availability, and real affiliate links across 50+ Singapore and Southeast Asia merchants.

The problem with shopping agents today

Building a shopping agent that works for more than one merchant means one of three things:

Platform scraping — maintain scrapers for every merchant, deal with proxies, CAPTCHAs, and rate limits. It works until it doesn't.
Official APIs — most merchants don't have developer-friendly APIs. The ones that do (Shopee, Lazada) require business registration and approval.
Browser automation — Playwright or Puppeteer for every product lookup. Slow, fragile, and expensive at scale.

None of these give you a clean JSON response with price, availability, brand, and affiliate link that a Claude agent can reason over.

What BuyWhere actually is

BuyWhere is a scraped, normalized, and continuously updated product catalog covering Singapore and Southeast Asia. Currently indexed: 1M+ products across 52 merchants, refreshed on a rolling schedule.

The API gives you:

Full-text product search with brand and category filters
Price comparison across merchants
30-day price history and trend signals
Deal detection (products with 10%+, 20%+, 30%+ discounts)
Affiliate-linked product URLs for Shopee, Lazada, and major SG retailers

Here is what a search looks like:

curl -s "https://api.buywhere.ai/v1/search?q=dyson%20vacuum&limit=5" \
  -H "Authorization: Bearer bw_live_xxxxx"

{
  "products": [
    {
      "id": "bw_prod_abc123",
      "title": "Dyson V15 Detect Absolute Vacuum Cleaner",
      "brand": "Dyson",
      "merchant": "Shopee",
      "price": 749.00,
      "currency": "SGD",
      "url": "https://shopee.sg/affiliate/redirect?...",
      "in_stock": true,
      "discount_pct": 12
    }
  ],
  "total": 47,
  "query_time_ms": 23
}

No HTML parsing. No proxy rotation. Just structured data.

The MCP server: agents that can actually shop

Raw API access is fine for developers. But the moment you want an AI agent to actually use the data, you need tool definitions — not just HTTP endpoints.

BuyWhere ships a Model Context Protocol (MCP) server that exposes seven tools:

Tool	What it does
`search_products`	Full-text search across all merchants
`compare_prices`	Products sorted by price for cross-platform comparison
`get_product`	Product details by BuyWhere ID
`get_deals`	Products with the biggest discounts (10%+)
`find_deals`	Best deals (20%+) with expiration dates
`browse_categories`	Category taxonomy with product counts
`get_category_products`	Paginated products within a category

Configure it once in your AI client and the agent can shop:

// Claude Desktop config (macOS)
{
  "mcpServers": {
    "buywhere": {
      "command": "python",
      "args": ["/path/to/mcp_server.py"],
      "env": {
        "BUYWHERE_API_KEY": "your_key",
        "BUYWHERE_API_URL": "https://api.buywhere.ai"
      }
    }
  }
}

Ask Claude: "Find the cheapest Nintendo Switch OLED deal in Singapore right now" — and it calls find_deals with product_name=nintendo switch oled, returns the best discount across all merchants, and gives you a clickable affiliate link.

Developer use cases we've seen

Deal alert agents — a Discord bot that watches for price drops on wishlisted products. Uses find_deals on a cron schedule, posts to a channel when a discount threshold is crossed.

Shopping concierge agents — a Claude agent that takes a budget and occasion ("birthday gift for my sister, $50, likes running") and returns 3-5 product options with price comparison, using search_products and compare_prices.

Price tracking dashboards — a Streamlit app that shows 30-day price history for a product across merchants, so users can decide whether to buy now or wait.

Affiliate monetization layers — a Cursor workspace where agents attach affiliate links to every product they recommend, converting recommendations into commission.

What we didn't build (yet)

User accounts and auth — the API is key-based. Personalization is on the roadmap.
Cart and checkout — we give you data and links. Conversion is on the merchant side.
Non-SG markets — we're Singapore-first. Thailand, Malaysia, and Philippines are on the roadmap.

Getting started

Docs: https://api.buywhere.ai/docs MCP guide: https://api.buywhere.ai/docs/guides/mcp GitHub: https://github.com/buywhere/buywhere-api

Free tier: 100 req/min. No credit card required.

What we want from you

If you're building shopping agents, deal bots, commerce copilots, or affiliate tools:

What endpoints are you missing?
What agent workflows still require browser automation?
What would make BuyWhere more useful than rolling your own scraper stack?

We're actively building what developers actually need. Comments, issues, and feature requests welcome.

BuyWhere is live at api.buywhere.ai. 1M+ products. 52 merchants. Structured for agents.