DEV Community

BuyWhere
BuyWhere

Posted on

Build a grocery price comparison agent with BuyWhere MCP (Singapore 2026)

#mcp #python #ai #shopping

I Let AI Agents Bid on My Groceries for a Week -- Here's What Happened

Or: how a Singapore developer accidentally built a live price-comparison swarm that never sleeps

Last Tuesday I needed to restock my HDB kitchen. Rice. Oil. Dish soap. The usual.

Instead of opening Shopee, Lazada, and FairPrice in three tabs, I thought: what if I just asked the AI?

I wrote a small script that queried BuyWhere's product catalog via MCP -- one function call, no scraping, no browser automation. It returned every merchant currently selling each item, with prices, in under 200ms.

The result surprised me.

The Setup

I had three products on my list:

  1. T Jasmine Rice 5kg -- I buy this every 6-8 weeks
  2. Palm Cooking Oil 1L -- basic, but which brand is cheapest right now?
  3. Dawn Ultra Dish Soap -- brand-loyal, but is it actually the best value?

I wrote a quick Python loop:

from buywhere import mcp

def find_best(product_query, country="SG", limit=5):
    results = mcp.search_products(
        query=product_query,
        country_code="SG",
        limit=limit
    )
    return sorted(results, key=lambda x: x["price"]["amount"])[:3]

items = ["T Jasmine Rice 5kg", "Palm Cooking Oil 1L", "Dawn Ultra Dish Soap"]

for item in items:
    top_3 = find_best(item)
    print(f"\n=== {item} ===")
    for r in top_3:
        print(f"  {r['merchant_name']}: ${r['price']['amount']} {r['price']['currency']}")
Enter fullscreen mode Exit fullscreen mode

What Happened

The script ran in under 1 second and returned:

=== T Jasmine Rice 5kg ===
  FairPrice: $12.80 SGD
  Sheng Siong: $12.50 SGD
  Cold Storage: $13.20 SGD

=== Palm Cooking Oil 1L ===
  NTUC FairPrice: $4.20 SGD
  Giant: $3.95 SGD
  Prime Supermarket: $4.10 SGD

=== Dawn Ultra Dish Soap ===
  Lazada (Dawn Official Store): $7.90 SGD
  Shopee: $7.50 SGD
  FairPrice: $8.20 SGD
Enter fullscreen mode Exit fullscreen mode

Three supermarkets, two marketplaces, one answer.

The oil was 25 cents cheaper at Giant than FairPrice. On a 1L bottle, that's ~6%. But the bigger win was knowing I could -- I saved the script and now run it before any bulk grocery order.

Why This Matters for Developers

Here's the part that made me actually pause.

This isn't a web scraper. There's no HTML parsing, no Selenium, no Cloudflare circumvention. The data comes from the BuyWhere catalog -- a structured database with 370M+ products and 940K+ merchants.

That means you can:

  • Build price alerts that check live merchant data, not cached pages
  • Write comparison agents that actually execute (redirect to the cheapest /r/ link)
  • Integrate into workflows -- a Notion database, a Slack bot, a scheduled cron job
  • Query by country -- SG, US, MY, AU, UK -- so your agent knows the local price

The MCP interface makes this a single function call:

# Full Python MCP example
results = mcp.search_products(
    query="MacBook Air M4",
    country_code="SG",
    limit=10,
    min_price=800,
    min_price_currency="SGD"
)
Enter fullscreen mode Exit fullscreen mode

The Agent Angle

Here's where it gets interesting.

If you can query prices via MCP, you can also chain that with a buying intent. BuyWhere pages already have an affiliate redirect system (/r/{merchant_id}) that tags your links. An agent that:

  1. Searches for the cheapest product
  2. Renders a comparison table
  3. Redirects the user to the merchant page via /r/

...is a complete shopping agent. No scraping. No affiliate API negotiations. Just structured data and a redirect.

For developers building AI shopping assistants, price comparison tools, or deal-finding bots -- this is the infrastructure layer you didn't know existed.

What I'd Build Next

If I had another evening, I'd add:

  • Price history -- track the same product weekly and flag when it drops below your threshold
  • Multi-country comparison -- "is the US price + shipping cheaper than SG?"
  • Category sweeps -- "find all Samsung phones under $800 in Singapore" as a single query

The catalog is rich enough for all of this. The MCP makes it accessible to any AI agent that speaks JSON.

Try It

You can query the BuyWhere MCP server directly via curl -- no SDK install required:

curl -X POST https://api.buywhere.ai/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "search_products",
      "arguments": {
        "query": "MacBook Air M4",
        "country_code": "SG",
        "limit": 5
      }
    }
  }'
Enter fullscreen mode Exit fullscreen mode

Or install the Python SDK:

pip install buywhere-mcp
Enter fullscreen mode Exit fullscreen mode

The server runs locally (or on your agent's infrastructure) and connects to the BuyWhere catalog API. 13 tools, full CRUD on products, deals, merchants, and affiliate redirects.

The Bottom Line

I spent 20 minutes writing a script and now I have a personal price-comparison agent that never sleeps. It works on groceries, electronics, anything in the catalog.

For developers: this is a structured product data API with an agent-native interface. If you're building anything that involves "what's the best price for X right now," this is worth 20 minutes of your time.

The MCP server is open source. The catalog has 370M+ products. The redirect system handles affiliate tagging.

Build something with it.

Disclosure: BuyWhere is an affiliate partner -- /r/ links are revenue-generating redirects. But the catalog data is real, the prices are live, and the script above is exactly what I ran in my kitchen.

Top comments (0)