Best Amazon APIs in 2026: 8 Tested and Compared

How We Evaluated

• What we tested: two requests on every provider. First, a product lookup by ASIN (B07ZPKBL9V, an Amazon Renewed iPhone 11, chosen because it has many third-party offers); what each provider returned for it is summarized in the table below. Second, a keyword search for "wireless earbuds", which is the request shown in each provider's code and response blocks. The latency sample adds four more ASINs (B0CHX3QBCH, B09G9FPHY6, B0BSHF7WHW, B08L5NP6NG) and two more search terms (standing desk, air purifier). We also called each provider's reviews and offers endpoints once to check they exist and respond; those results are in the coverage table. If you want the product-lookup code for these providers, our e-commerce API roundup shows it for seven of them.
• Pricing: every number was read from the provider's public pricing page or docs on September 4, 2026. Where a page contradicted another page on the same site, we say so. Monthly billing unless stated.
• Unit conversion: credits, results, records, tokens, and searches were converted to cost per 1,000 Amazon product-page requests, then to the monthly bill at 10K, 50K, and 200K requests.
• Code and responses: every code block was executed as printed, with only the placeholder key replaced, and every response block is trimmed from the JSON that call returned. Seven providers were tested on free accounts created for this article; SerpApi was tested with an existing OpenWeb Ninja account. The Creators API and Keepa sections use their official documentation, labeled, because both are gated: one by affiliate sales, the other by a paid plan.
• Response time: the same five ASINs and three searches, once each, from one machine in the same session. We report the median. It is a smoke test, not a benchmark.
• Order: by how many readers we expect each to fit, not strictly by price.

We build one of these APIs, so read the OpenWeb Ninja section with that in mind. The competitor sections are as flattering as the published numbers allow.

Best for: the widest set of Amazon endpoints at one flat per-request price

The Real-Time Amazon Data API documents 21 endpoints in its OpenAPI spec: search, products by category, product details, product reviews, review details, top reviews, product offers, seller profile, seller reviews, seller products, best sellers, deals, deal products, promo code details, influencer profile, influencer posts, influencer post products, ASIN to GTIN, GTIN to ASIN, category list, and scrape by URL, across 24 Amazon country codes. Every one of them costs one request. A reviews call or a seller lookup is billed exactly like a search page.

The same key also covers Walmart, eBay, Home Depot, and 40+ non-retail APIs. The free plan is 100 requests a month, no card. On reviews, the product page itself notes that Amazon's login wall limits results to 8 per call; our probe returned exactly 8 plus the rating distribution. Plans run from $25 for 10,000 requests to $2,000 for 5,000,000; the unit economics table below has the per-endpoint math.

Python example: keyword search

Response (trimmed)

28,266 total matches and 16 products on the first page, each with ASIN, price, original price, rating, review count, Prime and badge flags, and delivery text. Median 2.2 seconds across our three search terms; 3.9 seconds for product lookups.

Best for: cheapest parsed Amazon pages once you pass 100K a month

Oxylabs sells one Web Scraper API with per-site sources. For Amazon there are five dedicated sources, plus a generic amazon source for any Amazon URL: amazon_product, amazon_search, amazon_pricing (the offers list), amazon_sellers, and amazon_bestsellers, all with parse: true at no extra charge. There is no reviews source; our amazon_reviews attempt returned 400. Amazon pages are the cheapest bucket in their pricing: $0.50 per 1,000 on the $49 Micro plan down to $0.25 on Corporate.

The pricing page splits plans across a Regular tab (Free trial, Micro, Starter, Advanced) and an Enterprise tab (Venture, Business, Corporate, Custom+); a plain fetch of the page without JavaScript shows only five of them, which is worth knowing if you scrape pricing pages. Billing is per successfully fetched page, and a 2xx or 4xx counts even if the parser returned nothing useful.

Python example: keyword search

Response (trimmed)

Organic, paid, Amazon's Choice, and suggested results are separated, with 16 organic and 14 paid on this page, and prices are numbers. Median 2.5 seconds for search; 5.3 seconds for product lookups with one 25.8-second outlier.

Best for: the largest free tier, and reviews and seller datasets billed per record

Bright Data packages Amazon as 14 scrapers: products by URL or ASIN, by keyword, by category, by UPC, best sellers by category, reviews, seller info, product search, and global variants of each. You pay per delivered record: $1.50 per 1,000 on pay-as-you-go, $499 for 384,000 on Scale, and 5,000 free credits every month, shared across Bright Data's scraping products, no card. Output is JSON, NDJSON, JSON lines, or CSV.

The API is a dataset trigger, not a GET. You POST an array of inputs to a dataset ID; the synchronous endpoint has a one-minute window, after which you get a 202 with a snapshot ID to poll. Keyword search is a discovery dataset that returns one record per product found, so a 16-result page is 16 records, not one request. That is why we show the product scraper here rather than search: it is the only endpoint whose cost maps to one call.

Python example: product by URL

Response (trimmed)

99 fields per record, numeric prices, and a seller ID you can chain into the seller scraper. Median 9.0 seconds for product lookups, the slowest of the eight; in our two days of testing, five of six synchronous calls returned inline and one came back as a snapshot to poll.

Best for: teams on ScraperAPI for general scraping who want parsed Amazon pages from the same credits

ScraperAPI is a general proxy-and-render scraper with structured-data endpoints layered on top. For Amazon there are three: product, search, and offers, each with an async variant. Any Amazon structured request costs 5 credits, so the $49 Hobby plan's 100,000 credits are 20,000 Amazon pages, $2.45 per 1,000. That drops to $0.745 at Startup and under $0.50 from Business up. Only 200 and 404 responses are charged.

There is no reviews endpoint; our call to a review path returned 404. The offers endpoint worked. ScraperAPI and Rainforest share a parent company but remain separate products with separate pricing.

Python example: keyword search

Response (trimmed)

18 results with both a price string and a numeric price, plus next-page URLs. Median 3.1 seconds for search with one 14-second outlier; 3.2 seconds for product lookups, tied with Scrapingdog for fastest.

Amazon Unit Economics: Cost per 1,000 Requests by Endpoint

Published pricing on September 4, 2026, converted to cost per 1,000 requests of each Amazon endpoint type. The first figure is the entry paid plan, the second is the cheapest listed self-serve tier. Where a provider bills per record or per result, a search page counts as one request only if the provider bills it that way. Full plan ladders for seven of these providers are in the e-commerce roundup.

Read across, not down. Scrapingdog is the cheapest product page here and among the most expensive reviews page, because reviews cost five credits and do not currently work. Rainforest and OpenWeb Ninja charge the same for every endpoint, which is why their columns are flat. Bright Data and Apify look cheap per unit until a search returns 40 units. Keepa's per-request figures assume you use the full 892,800 monthly tokens; the plan costs 49 EUR whether you do or not.

Measured Response Time (Same 5 ASINs, Same 3 Searches):

One request to each provider's Amazon product endpoint per ASIN (B07ZPKBL9V, B0CHX3QBCH, B09G9FPHY6, B0BSHF7WHW, B08L5NP6NG), and one request to each provider's Amazon search endpoint per term (wireless earbuds, standing desk, air purifier), on September 4, 2026, from a single machine in Israel. Every call succeeded. Small sample; it tells you the order of magnitude, not the tail.

Field counts are top-level keys in the product object for ASIN B07ZPKBL9V. More fields is not automatically better; check that the ones you need are populated. Apify search returned 10 items per run, so its time covers a crawl, not a single page.

The Official Route: Amazon's Creators API (PA-API 5 Is Deprecated)

If you search for "Amazon product API" you will find guides to Product Advertising API 5.0. As of September 2026 it is gone. Every PA-API 5 documentation URL we tried redirected to a deprecation notice, which states that calls now return HTTP 403 with AccessDeniedException and the message "Product Advertising API is deprecated. Please migrate to Creators API using the migration guide." The successor is the Creators API: REST, OAuth 2.0 client credentials, lowerCamelCase fields.

Eligibility is the catch. The docs require an approved Amazon Associates account with "at least 10 qualifying sales within the past 30 days", and access is revoked after 30 consecutive days without qualified referring sales. New credentials get 1 request per second and 8,640 per day; limits grow by one request per day for every five cents of shipped revenue you refer, up to 10 requests per second. Amazon's docs list no fee.

What you do not get: customer reviews (no review resource exists), third-party offer listings (OffersV2 "provides only featured listings as shown to retail customers on Amazon Detail Page buybox"), or any history. Items can also come back as "not accessible through the Creators API." That gap is why the eight scraping APIs above exist.

Request (from the Creators API docs, verbatim)

Response (from the getItems docs page, trimmed)

Copied from the getItems reference and the cURL guide. We did not run these calls: the Creators API requires an Associates account with recent qualifying sales, which a research account does not have. Price arrives under an offersV2.listings.price resource as money.amount, currency, and displayAmount.

How to Choose: 4 Common Scenarios