To price watch on Amazon, open the product page and tap the price history link next to the price, or ask the shopping assistant "what's the price history?". For ongoing alerts, add the item to a wishlist or use a tracker extension. For hundreds of products, poll the listings yourself with a scraping API and set your own thresholds.
What price watching on Amazon actually means
The phrase covers three jobs that look similar and need different tools:
- One-off sanity check. You are about to buy something and want to know whether $248 is normal or a spike. Amazon's own price history answers this in two taps.
- Standing watch on a few items. You want a nudge when a specific pair of headphones drops below $199. Wishlists, deal alerts and third-party trackers cover this.
- Systematic monitoring. You care about 200, 2,000 or 20,000 listings — your own catalogue, a competitor's, or a category you are researching — and you need the numbers as rows in a table, not as push notifications. That is a data problem, and it needs an API.
One thing to internalise before you build anything: an Amazon price is not a single number. The buy box rotates between sellers, coupons and Prime-exclusive discounts stack on top of the list price, variants of the same parent ASIN price differently, and the figure you see depends on the marketplace and the country you are browsing from. Any watcher that records one scalar per ASIN and ignores seller, coupon and variant will produce alerts you cannot trust.
Option 1: Amazon's built-in price history and alerts
Price history on the product page
Amazon exposes price history natively through its shopping assistant. There are two entry points: a price history link right next to the price on eligible product detail pages, or the assistant icon (bottom right in the Amazon Shopping app, top navigation on desktop), where you can ask "has this item been on sale in the past 30 days?" or "is this the lowest price recently?". Amazon says the view now spans 30, 90 and 365 days, that over 50 million customers have used it since it launched in 2024, and that the average user checks it three times a month; availability is US, UK, Canada and India, with 365-day insights rolling out in the US, UK and India (About Amazon). The assistant formerly called Rufus was renamed Alexa for Shopping on 13 May 2026.
This is the fastest answer to "is this Prime Day price actually a deal" and it costs nothing. Its limits are equally clear: it is per-product, read-only, human-paced, and it will not hand you a CSV.
Amazon price alert wishlist and deal alerts
The closest thing to a native alert is a list. Add items to a wishlist and Amazon surfaces price-drop badges against them, and the app can push deal notifications for items and deals you have watched. In the Amazon app the switches live under your account settings in the notifications section — exact menu labels shift between app versions, so if you are trying to manage price alerts on Amazon and the path in a guide does not match, search the settings screen for "notifications" rather than following screenshots.
Two caveats people discover the hard way. First, a wishlist drop badge is not a threshold: you cannot say "tell me at $199", only "tell me it moved". Second, notification delivery is best-effort — fine for a coat, unreliable if you are chasing a doorbuster.
Auto buy, and how to cancel auto buy on Amazon
Amazon's agentic shopping features include auto-buy, where you set a target price and Amazon places the order if the price reaches it, alongside things like handwritten shopping list transcription and personalised shopping guides. Treat it as a live instruction, not a reminder. Manage or cancel it from the place you created it — the product page or your list of active auto-buy requests in the shopping assistant — and do it before the trigger fires, because cancelling the watch does not unwind an order that has already been placed. If an order has gone through, you are in ordinary order-cancellation and returns territory instead.
Option 2: trackers and extensions — evaluate, don't rank
Dedicated Amazon price trackers and browser extensions exist for exactly the gap the native features leave: long charts, target-price emails, and cross-marketplace views. Rather than rank them, judge any candidate against criteria that predict whether you will still trust it in six months:
- History depth and provenance. How far back does the chart go, and is it their own observations or an inference? Ask what happens to a gap in the series.
- Granularity. Does it track the specific variant and the specific seller, or collapse a parent ASIN into one line? Does it separate buy-box price from third-party and warehouse offers?
- Alert semantics. Absolute target, percentage drop, or lowest-in-N-days? Can you express "below trailing 90-day median"?
- Latency. Check interval and notification channel. Hourly email is a different product from a five-minute webhook.
- Marketplace coverage. Prices differ per country; a US-only tracker is useless for a UK buyer.
- Export. Can you get the raw series out as CSV or via an API, or is it locked in a dashboard?
- What the extension can read. A price alert extension that requests access to every site you visit is a browsing-history question, not just a shopping one.
- How it is funded. Most are affiliate-funded. That is fine, but it means the incentive is clicks to buy, not neutrality.
For personal shopping, a good tracker beats anything you would build. The moment you need the data — for pricing decisions, category research, or feeding a model — the calculus flips.
Option 3: run your own price watch with an API
A self-hosted watcher is a short loop: resolve a list of listing URLs, fetch each one, extract price and availability, store the observation with a timestamp, compare against a rule, act. The reason to own it is that you keep the time series, you choose the cadence, and the same pipeline extends beyond Amazon to any retailer — the pattern we describe in how to price monitor and productise in competitor price monitoring.
Step by step
- Build the watchlist. Canonical
/dp/ASINURLs per marketplace domain. If you do not have ASINs, discover them with search — a SERP API call against the shopping vertical returns candidate listings you can dedupe into a list. - Decide the fields. At minimum: price, currency, list/strike price, coupon, buy-box seller, availability text, and the variant identifier. Record nulls explicitly; a missing price is signal, not noise.
- Choose the fetch mode. Try plain HTTP first and only turn on JS rendering for pages where the number genuinely is not in the initial HTML. Route requests through residential proxies with country targeting so you see the price a local shopper sees, not a geo-mismatched one.
- Store every observation. Timestamp, source URL, raw markdown or HTML snapshot, parsed fields. That table is your price history, and it is the part no tracker will give you.
- Write the rule. "Below $199" is the beginner version. "Below the trailing 90-day median by 10% and in stock from the brand's own storefront" is the version that does not page you at 3am for a third-party marketplace typo.
- Tier the cadence. Hot items hourly, the long tail daily or weekly. Cadence, not scale, drives your bill.
- Deliver and audit. Webhook or email, plus a link back to the stored snapshot so a human can verify the claim.
Fetching one listing
The Web Scraping API returns a page as clean Markdown from $0.0002, or $0.001 with JS rendering, in a single envelope:
curl https://api.quanticdata.io/v1/scrape \
-H "Authorization: Bearer $QD_API_KEY" \
-d '{ "url": "https://www.amazon.co.uk/dp/B0EXAMPLE1" }'
{ "success": true,
"data": { "markdown": "# Product title\n£248.00 …" },
"usage": { "cost_usd": 0.0002 } }Every endpoint answers with the same success / data / usage shape, and a failed call returns success: false with an error code and costs nothing. Parameter names for CSS and AI extraction live in the documentation.
Sweeping a whole watchlist
For known URLs, batch is the right primitive: up to 1,000 URLs per job at $0.0002 each, charged up front on requested volume with the unfetched share auto-refunded when the job settles.
import os, time, requests
H = {"Authorization": f"Bearer {os.environ['QD_API_KEY']}"}
BASE = "https://api.quanticdata.io/v1"
asins = ["B0EXAMPLE1", "B0EXAMPLE2", "B0EXAMPLE3"]
urls = [f"https://www.amazon.com/dp/{a}" for a in asins]
job = requests.post(f"{BASE}/batch", headers=H,
json={"urls": urls}).json()
job_id = job["data"]["jobId"]
while True:
r = requests.get(f"{BASE}/batch/{job_id}", headers=H).json()
if r["data"]["status"] in ("completed", "failed"):
break
time.sleep(5)
for page in r["data"]["results"]:
price = parse_price(page["markdown"]) # your parser
save_observation(page["url"], price) # url + price + now()
if price and price <= target[page["url"]]:
notify(page["url"], price)
Keep parse_price boring and defensive: reject values outside a plausible band for that ASIN, and never overwrite yesterday's row — append.
Honest cost math
Watching prices is cheap; watching them too often is not. At list prices of $0.0002 per plain page and $0.001 with JS rendering:
| Watchlist and cadence | Pages / month | Plain HTTP | JS rendered |
|---|---|---|---|
| 25 items, hourly | 18,000 | $3.60 | $18.00 |
| 500 items, 4× a day | 60,000 | $12.00 | $60.00 |
| 5,000 items, daily | 150,000 | $30.00 | $150.00 |
Two conclusions. First, rendering is a 5× multiplier — spend an afternoon proving which pages actually need it. Second, the pay-as-you-go tier includes $2 of free usage every month with no card, which is roughly 10,000 plain page fetches: enough to prototype a personal watcher without a subscription. Compare that with any per-seat dashboard before you sign anything, and note that with pay-per-success the pages that get blocked do not appear on the bill.
Agent workflows: price watching from inside your assistant
If you already work in Claude, Claude Code, Cursor, Windsurf, VS Code or Cline, you can skip the cron job for exploratory work. Our MCP server exposes eight tools — search, scrape, map, crawl, crawl_status, batch, batch_status, seo_audit — so an assistant can resolve a product from a description, fetch the current listings, and tell you where today's price sits relative to what it recorded last week. Residential proxies sit underneath, billing is per success, and the free open package plus $2 free per month covers a lot of one-off questions.
That pattern generalises. The same search-then-scrape loop that answers "is this a deal" also builds category-level pricing panels, which is why it sits at the centre of our market research data work: every row arrives with a source URL and a timestamp, so a price claim is auditable months later. If you want the conceptual background on how agents consume these tools, is an MCP server like an API? covers the difference between a tool call and a plain endpoint.
Staying inside the lines
None of this is legal advice. A few practical boundaries worth setting yourself before you write a scheduler:
- Read the platform's terms. Amazon's Conditions of Use restrict automated access, data mining and robots, and terms are a contract question that is separate from copyright or computer-misuse law.
- Stay on public pages. Anything behind a login, a paywall or an account you do not own is a materially different risk profile.
- Collect facts, not people. Prices, availability and titles are commercial facts; reviewer names and profiles are personal data with their own rules.
- Be gentle. Tiered cadence, backoff on errors, and caching so you never fetch the same page twice in a minute.
- Prefer official interfaces where they exist. If you are a seller or an affiliate, Amazon's own APIs are the sanctioned path and generally cheaper to maintain.
For a fuller treatment of where the lines fall, see is AI web scraping legal?. If in doubt about a commercial monitoring programme, ask a lawyer in your jurisdiction rather than a blog.
Picking the right option
Buying one thing this week: use the native price history, then set an auto-buy or wishlist watch and move on. Tracking a dozen items across marketplaces: a dedicated tracker, chosen on history depth, alert semantics and export. Turning Amazon prices into an input for pricing, sourcing or research decisions: own the loop, store every observation, and pay per page fetched. The three approaches are not competitors — most teams end up using the first for shopping and the third for work.