Amazon Product Scraper (Apify actor)

Scrapes Amazon product data based on URL and country, without the Amazon API: reviews, prices, descriptions, ASINs, ratings, images, Best Sellers Rank, and the full product-information tables.

Feed it any mix of Amazon URLs — search results, category / browse-node pages, Best Sellers pages, or product detail pages — and/or a bare search keyword plus a country, and it writes one JSON record per product to the run's dataset.

Input

Field	Type	Default	Notes
categoryOrProductUrls	[{ url }]	—	Search (/s?k=…), category (/b?node=…), Best Sellers (/gp/bestsellers/…), or product (/dp/ASIN) URLs. Bare ASINs also accepted. Aliases: startUrls, productUrls.
search	string	—	Bare keyword; searched on the country marketplace. Aliases: keyword, searchKeywords, keywords, queries.
country	string	US	Marketplace for bare keywords + proxy geolocation. 23 marketplaces supported (US, GB/UK, DE, FR, IT, ES, CA, JP, IN, MX, BR, AU, NL, SE, PL, TR, AE, SG, SA, BE, IE, EG, ZA). A URL's own domain always wins — country then only steers the proxy.
maxItems	int	100	Total product cap for the run. Alias: maxResults.
maxItemsPerStartUrl	int	0 (off)	Per-start-URL cap. Alias: maxProductsPerStartUrl.
maxSearchPagesPerStartUrl	int	5	Result pages walked per search/category URL (Amazon serves ≤ ~20).
scrapeProductDetails	bool	true	OFF = shallow, fast listing-only records (no dp-page visits).
scrapeReviews	bool	true	Alias: includeReviews. Requires details ON.
maxReviews	int	20	Per product. Amazon caps anonymous review pages at ~10 (~100 reviews); where the reviews pages are login-walled (increasingly common since late 2024) the actor keeps the product page's top reviews instead.
reviewsSort	helpful|recent	helpful	Review page order.
proxyConfiguration	proxy	Apify RESIDENTIAL	Amazon blocks datacenter IPs quickly — keep residential.

Robot-check pages are detected (captcha form, block-page markers) and retried on a fresh session/IP automatically; requests that still fail salvage what's already known (listing-row data, on-page reviews) instead of dropping the product.

Output (one record per product)

{
  "asin": "B0TESTMUG1",
  "url": "https://www.amazon.com/dp/B0TESTMUG1",
  "title": "Bigfoot Sasquatch Coffee Mug, 15 oz Ceramic",
  "brand": "CryptidWorks",
  "price": 18.99,
  "priceRaw": "$18.99",
  "currency": "USD",
  "listPrice": 24.99,
  "rating": 4.6,
  "reviewsCount": 1234,
  "inStock": true,
  "availability": "In Stock",
  "featureBullets": ["15 oz ceramic mug…", "Dishwasher and microwave safe"],
  "description": "Start every morning with the legend himself…",
  "image": "https://m.media-amazon.com/images/I/71test._AC_SL1500_.jpg",
  "images": ["…hi-res first…"],
  "breadcrumbs": ["Home & Kitchen", "Mugs"],
  "bsr": 12345,
  "bestsellerRanks": [{ "rank": 12345, "category": "Home & Kitchen" }, { "rank": 678, "category": "Coffee Mugs" }],
  "attributes": { "Brand": "CryptidWorks", "Material": "Ceramic", "ASIN": "B0TESTMUG1" },
  "reviews": [
    {
      "id": "R1TEASER001", "title": "Best mug I own", "rating": 5,
      "body": "Sturdy, funny…", "author": "Jane D.",
      "date": "March 3, 2025", "country": "United States",
      "verified": true, "helpfulVotes": 12, "variant": "Color: Forest Green"
    }
  ],
  "sponsored": false,
  "domain": "amazon.com",
  "country": "US",
  "scrapedAt": "2026-07-05T12:00:00.000Z"
}

The field names (asin, title, brand, price, rating, reviewsCount, bsr, url, image) line up with what the common Amazon actors emit, so consumers built for those (including the Suppliers app's src/ai/amazon-scan.ts normalizer) read this actor's output unchanged.

Develop / test / deploy

npm install
npm test                    # fixture-pinned parser tests, no network

# local run against real Amazon (uses your logged-in `apify` CLI for proxy):
apify run --purge --input '{"search":"bigfoot mug","country":"US","maxItems":5,"maxReviews":5}'

# deploy to your Apify account:
apify login
apify push

Once pushed, point the Suppliers app at it by setting APIFY_AMAZON_ACTOR=<your-apify-username>/amazon-product-scraper on Render (the actor name comes from .actor/actor.json; the app's default is a third-party store actor — this makes the Amazon market scan run on your own actor and your own proxy budget).

Design notes

• CheerioCrawler, no browser — every page shape used (search grid, dp page, reviews pages, bestseller grid) renders its data server-side, so plain HTTP is ~10× cheaper than a headless browser. Bestseller grids lazy-load beyond ~30 items per page; the actor takes both static pages (pg=1,2) which covers the Top-100 lists' server-rendered portion.
• Fallback selector chains everywhere — Amazon A/B-tests layouts; every field reads primary + legacy selectors (test/fixtures/ pins them).
• Reviews strategy — dp-page top reviews are captured first (always available), then the dedicated /product-reviews/ pages are walked up to maxReviews / Amazon's ~10-page anonymous cap; a login-walled reviews page degrades gracefully to the top reviews rather than losing the product.
• Budgeting — a scheduled counter reserves dataset slots at listing time so the crawler never fans out to hundreds of dp pages it won't use; maxItems triggers a graceful crawler.stop().