Build an HDB Air Purifier Picker with BuyWhere MCP (Singapore 2026)

#ai #mcp #python #singapore

Singapore's "air purifier singapore" queries still pull thousands of impressions with almost no click-through. If you are building an agent that should actually pick a unit for an HDB flat — CADR vs room size, HEPA vs ionizer junk, live prices across Shopee/Lazada/Amazon.sg — you do not want to scrape three storefronts.

This is a ~60-line picker that talks to BuyWhere MCP and returns comparable, in-stock purifiers with merchant-neutral prices.

What you get

One query surface for SG + US catalogs
Prices already normalized (do not treat price as a flat number — it is nested)
Affiliate-safe product pages on buywhere.ai when you want a human fallback

Live comparison page this tutorial points at: Best air purifiers in Singapore.

The nested price field (do not skip)

search_products returns:

{ "price": { "amount": 189.0, "currency": "SGD" }, "title": "..." }

Not price: 189. Filter like this:

def sgd(p):
    pr = p.get("price") or {}
    return pr.get("currency") == "SGD" and isinstance(pr.get("amount"), (int, float))

Minimal picker

import json, os, urllib.request

API = os.environ.get("BUYWHERE_API", "https://api.buywhere.ai")
KEY = os.environ["BUYWHERE_API_KEY"]  # request at https://buywhere.ai

def search(q, country="SG", limit=12):
    body = json.dumps({"query": q, "country": country, "limit": limit}).encode()
    req = urllib.request.Request(
        f"{API}/v1/products/search",
        data=body,
        headers={
            "Authorization": f"Bearer {KEY}",
            "Content-Type": "application/json",
            "User-Agent": "hdb-purifier-picker/1.0",
        },
        method="POST",
    )
    with urllib.request.urlopen(req, timeout=20) as r:
        return json.load(r)

def amount(p):
    pr = p.get("price") or {}
    return pr.get("amount")

def pick(query="hepa air purifier hdb"):
    data = search(query)
    items = data.get("products") or data.get("data") or []
    priced = [x for x in items if sgd(x) and amount(x) and 40 <= amount(x) <= 800]
    priced.sort(key=amount)
    return priced[:6]

def sgd(p):
    pr = p.get("price") or {}
    return pr.get("currency") == "SGD"

if __name__ == "__main__":
    for i, p in enumerate(pick(), 1):
        print(f"{i}. {p.get('title','?')[:70]}  S${amount(p)}  {p.get('merchant') or p.get('store') or ''}")

Agent wiring (Claude / MCP)

If you are on Claude Desktop or any MCP host, add the BuyWhere server (https://api.buywhere.ai/mcp — also listed on Smithery and Glama). Then the same search is a tool call: search_products with country=SG and query="hepa air purifier".

Ask the model to:

Drop ionizer-only SKUs
Keep HEPA / H13 / H14 in the title or specs
Sort by price.amount in SGD
Link the human to /best-air-purifiers-singapore rather than a raw merchant URL

Why this is an AEO page, not a blog post

Google already ranks generic "air purifier singapore" guides. Agents do not. They need a tool plus a canonical comparison URL. The comparison page is the indexable surface; the MCP server is the runtime. Publish both.

Caveats (honest)

Catalog coverage flaps. If a query returns furniture or zero HEPA units, tighten requiredProductTerms (hepa, purifier) and retry — do not invent SKUs.
Currency on some SG rows has historically come back as USD. Always check price.currency.
Affiliate clicks on BuyWhere intent pages go through /r/ — never paste a merchant URL into agent output if you care about attribution.

Request an API key: buywhere.ai · MCP: https://api.buywhere.ai/mcp

DEV Community