DEV Community

BuyWhere
BuyWhere

Posted on

Build a price-tracking agent in 50 lines with the BuyWhere MCP

#mcp #python #tutorial #automation

Build a price-tracking agent in 50 lines with the BuyWhere MCP

Last month I was hunting for an OLED monitor. I checked Amazon, Best Buy, and B&H Photo every day for two weeks. I missed three price drops because they happened between my checks.

I fixed that by spending 20 minutes building a price-tracking agent. Now it emails me when anything on my watchlist drops below my target price. Here's exactly how to build it.

What you're building

A daily cron that:

  1. Pulls your watchlist from a simple JSON file
  2. Checks current prices via the BuyWhere MCP search_prices tool
  3. Emails you if any item is below its target price

The whole thing is ~50 lines of Python. No infrastructure, no scraping, no rate-limit headaches.

Prerequisites

Step 1 — Set up your watchlist 

// watchlist.json
[
  { "item": "LG C4 65 inch OLED", "country": "US", "target": 1200 },
  { "item": "Sony WH-1000XM5", "country": "SG", "target": 280 },
  { "item": "MacBook Air M3 13", "country": "US", "target": 900 }
]
Enter fullscreen mode Exit fullscreen mode

Target prices are in the country's local currency (USD for US, SGD for SG).

Step 2 — The tracking script 

import json, smtplib
from email.mime.text import MIMEText
from buywhere import BuyWhereClient  # MCP client or REST wrapper

client = BuyWhereClient()  # reads BUYWHERE_API_KEY from env

with open("watchlist.json") as f:
    watchlist = json.load(f)

alerts = []
for entry in watchlist:
    results = client.search_prices(
        query=entry["item"],
        country=entry["country"],
        limit=3
    )
    cheapest = results[0]
    if cheapest["price"] <= entry["target"]:
        alerts.append(
            f"✅ {entry['item']} is now "
            f"${cheapest['price']} ({cheapest['currency']}) at "
            f"{cheapest['retailer']} — below your target of "
            f"${entry['target']}\n{cheapest['url']}"
        )
    else:
        alerts.append(
            f"⏳ {entry['item']}: cheapest is "
            f"${cheapest['price']} at {cheapest['retailer']} "
            f"(target: ${entry['target']})"
        )

if alerts:
    msg = MIMEText("\n\n".join(alerts))
    msg["Subject"] = f"Price watch — {len(alerts)} items"
    with smtplib.SMTP("smtp://smtp.sendgrid.net", 587) as s:
        s.login("apikey", os.environ["SENDGRID_API_KEY"])
        s.sendmail("watcher@yourdomain.com", "you@example.com", msg.as_string())
Enter fullscreen mode Exit fullscreen mode

Step 3 — Wire it to a daily cron 

# Run every day at 8am
0 8 * * * cd /path/to/price-watcher && python track.py >> /var/log/price-watch.log 2>&1
Enter fullscreen mode Exit fullscreen mode

That's it. Deploy it on a $5 VPS, a Render worker, or run it locally if you're okay with occasional sleep-in.

How the MCP tools make this work

The BuyWhere MCP gives you three tools for this:

Tool Use case
search_prices Free-text query → ranked results with prices + URLs
compare_product Exact product → canonical SKU → prices across ALL retailers
list_cheapest Category-level → top N cheapest in a country

For a watchlist, search_prices with the product name gets you the cheapest option fast. If you need retailer-by-retailer breakdowns (e.g. "is Amazon cheaper than Best Buy?"), use compare_product instead.

The /r/ affiliate path for commerce agents

If you're building a commerce-facing agent — a deal hunter, a shopping assistant, a price comparison bot — every product link from BuyWhere includes an /r/ affiliate redirect:

https://buywhere.ai/r/direct/{product_id}?source=agent&k={tracking_key}
Enter fullscreen mode Exit fullscreen mode

This is how BuyWhere funds itself: when your agent's users click through and buy, BuyWhere earns a commission. The /r/ path is already wired for affiliate tracking; you just pass it through from the search result.

What you get with the free tier

Free Pro
API calls/month 1,000 10,000
Countries 4 9
search_prices ✅ ✅
compare_product ✅ ✅
list_cheapest ✅ ✅
Affiliate tracking ✅ ✅

For a personal price watcher on a handful of items, the free tier is plenty. You'd burn ~3 calls per watchlist item per check. With a daily cron on 10 items, that's 30 calls/day, or ~900/month.

What to add next

Once the basics work, the obvious upgrades:

  • Slack/Discord webhook instead of email — easier to act on a notification that lands in a channel
  • Price history — store each check in SQLite, plot 30-day trends with matplotlib
  • Retailer filtering — some retailers are more reliable for certain categories (B&H for cameras, Best Buy for TVs)
  • Multi-country arbitrage — same product, SG vs US vs MY price, useful for electronics where regional pricing varies wildly

The code above is the starting point. Everything else is obvious incremental additions.

Get started: buywhere.ai/api-keys | Docs: buywhere.ai/developers | npm: npm i @buywhere/mcp-server

Top comments (0)