Amazon Search Scraper

Collect Amazon search cards from keywords or marketplace search URLs, with each card's displayed price, rating, review count, rank, badges and delivery information.

{"keywords": ["wireless headphones"], "marketplace": "US", "maxPages": 1, "maxItems": 16}

Supply keywords, searchUrls, or both. Omitted or empty query arrays collect no rows and incur no row charges; the start charge still applies. Each keyword and URL is a separate query; repeated ASINs across queries or pages remain separate rows.

Input

Field	Type	Default	What it does
keywords	array	None	Keyword searches; the input form prefills wireless headphones.
searchUrls	array	None	HTTPS Amazon marketplace /s URLs; preserves department and refinement parameters.
marketplace	string	US	Keyword host: US, UK, DE, JP, IN, AU, IT, ES, MX, BR, NL, SE, PL, TR, SG, SA, AE or EG. URLs use their own host.
language	string	Omitted	US: en_US/es_US; DE: en_GB/de_DE; JP: en_US/ja_JP; IN: en_IN/hi_IN; IT: en_GB/it_IT; ES: en_GB/es_ES. Other host/value combinations are refused.
sort	string	Source choice	relevanceblender, price-asc-rank, price-desc-rank, review-rank, date-desc-rank or exact-aware-popularity-rank.
minPrice	number	Omitted	Amazon's minimum-price parameter in major currency units.
maxPrice	number	Omitted	Amazon's maximum-price parameter in major currency units.
minStars	number	Omitted	4 follows the page's own four-star refinement; other values are refused.
primeOnly	boolean	false	true is unavailable: the logged-out pages expose no Prime refinement.
maxPages	integer	1	Maximum pages per query, following the source Next link; at most 20.
maxItems	integer	16	Maximum delivered rows per query.
maxResults	integer	1000	Maximum delivered rows across the run.
includeSponsored	boolean	true	false omits sponsored cards after source ranks are assigned.
deliveryZip	string	Omitted	Anonymous delivery ZIP/postal code within the marketplace country.
deliveryCountry	string	Omitted	Two-letter country selected from the site's anonymous delivery options.
proxyConfiguration	object	Residential by marketplace	Optional buyer proxy settings. The form prefill selects RESIDENTIAL; country is derived from the target host unless supplied.

Zero for any row or page limit collects no rows; the start charge still applies. Price, star and sort filters use Amazon's own parameters, and unsupported source options are refused. Amazon's price filter can include edge rows; the Actor preserves the source answer. Queries end when the source has no Next link, a limit binds, or the collection time budget is reached. Pagination varies by query; a source can end before the requested page limit.

Output

One row represents one complete source card occurrence. Errors are saved in ERRORS, with query coverage and transport outcomes in SUMMARY, and never enter the billable dataset. A source-confirmed no-results page is a successful empty answer.

Fields	Meaning
query, searchUrl, queryServed	Requested query and starting URL, plus the page's own query or spelling substitution.
marketplace, language, scrapedAt	Verified page host code, actual page language and UTC collection timestamp.
page, rank, positionOverall	One-based source page/card ranks. Overall position adds preceding pages' card counts for a walk starting on page one; it is null throughout a walk starting later. Sponsored cards count before removal; limits never renumber rows.
asin, title, url, imageUrl	The card's own identity, title, product link and image.
price, listPrice	Displayed money as {value, currency, raw}; value is in major currency units, and raw retains the source text. List price is the displayed strike-through price.
priceType	featured, other_offer, subscription_offer, or null when no price is displayed.
rating, reviewCount	Individual product rating on the source's five-star scale and integer review count; brand averages and purchase counts are excluded.
isSponsored, isPrime, isBestSeller, isAmazonChoice	Card-owned flags. Prime is null when the card offers no delivery or Prime marker.
boughtPastMonth, offersCount, couponText, dealText	Displayed purchase count, offer-count text, coupon and deal text.
deliveryText, fastestDeliveryText, locationText	Card delivery text and the page's delivery location.
variantTeaser, sustainability	Displayed variant and sustainability text.

A value the card does not display stays null. A displayed value that cannot be extracted becomes an unbilled extraction error; it is never replaced with a guessed value. Currency comes from the displayed marker within the page's verified marketplace identity. Explicit displayed currencies are preserved even when different from that marketplace's home currency. In two 16-row amazon.com runs on 2026-10-12, all 32 rows carried USD prices and US delivery-location text. Lazy carousel placeholders emit no row.

Access and limits

The primary path uses anonymous browserless requests through residential proxies. A slow first read can start one more fresh identity after 0.75 seconds; the accepted identity is kept for later reads and the other is retired, and a read still in flight is cancelled. This can add proxy traffic and does not guarantee faster small jobs. SUMMARY counts the extra requests and their known completed transfer bytes; cancelled reads have unknown byte totals.

Classified source refusals rotate to a fresh identity within a bounded ladder. WAF, captcha, login and rate-limit responses stop new primary requests and switch to the shared browser service, unless an already-running first read returns an accepted answer. Browser fallback is capped: reads beyond 3 browser fallbacks per run are skipped, listed in ERRORS and not charged. Buyer proxy settings are passed to the browser where supported; unsupported custom URLs, groups or countries produce an unbilled fallback error.

Collection stops about 30 seconds before the run's timeout (about 270 seconds at the default 300-second timeout), and a charge limit stops further collection. No product-detail, offers, sellers or review-list pages are crawled.

CA is available through the browser fallback at higher cost; not yet offered. CA keywords and URLs are refused before collection. FR is unavailable: search returns HTTP 202 through every route measured; its keywords and URLs are also refused. A requested delivery location is refused if the browser service cannot preserve it. The six listed language preference pairs use the source's language parameter; other values are refused. UK and AU preference pages expose no language selector, and MX exposes no language-preference link on the measured search page. Additional marketplace language preferences have not been verified. Prime-only filtering remains unavailable on the logged-out source.

Amazon's Conditions of Use exclude commercial use of product listings, descriptions and prices, and data-mining or similar extraction tools. This restriction applies to the source; buyers should assess their intended use. Search cards and availability change over time and with the delivery location.

Billing

Each returned row is one result event. Each run also has a start event, charged once per GB of run memory (one event at the default 256 MB); a run that returns no rows still pays it. Records in ERRORS are never charged.