Two ways to pull Amazon product data, in the order most people actually need them. Lane one is a no code Amazon scraper built from Make and Airtable, with no repository to clone and nothing to deploy. Lane two is the dedicated Amazon endpoints on ScrapingBee's web scraping API, for the point where a spreadsheet stops being enough.
Both lanes read the same public Amazon pages. The difference is who runs the loop.
Everything below was executed against the live API on 2026-09-10. Every response field, every parameter name and every credit figure in this README came back from a real call, not from a documentation page.
The no code lane is five moving parts and no code at all:
- A Make scenario, triggered on a schedule.
- A ScrapingBee module inside it, set to Make an API call, method
GET. - An Amazon search URL in the URL field.
- Extraction rules, pasted into Extract Rules (JSON) under Show advanced settings.
- An Iterator, then an Airtable Create a Record action, so one scraped product becomes one row.
The extraction rules are the only part that looks like programming, and they are just a description of the page:
{
"results": {
"selector": "div.s-result-item.s-asin",
"type": "list",
"output": {
"link": { "selector": "a.a-link-normal", "output": "@href" },
"name": "span.a-size-medium.a-color-base.a-text-normal",
"price": "span.a-offscreen"
}
}
}Leave Render JS off. Amazon search results are in the delivered HTML and rendering costs extra credits for nothing. In the Airtable step, parseNumber(price; ".") strips the currency symbol, and the link values are relative, so prefix https://www.amazon.com before writing the row.
The Make integration is the version documented here. n8n and Zapier expose the same call if your automation already lives somewhere else.
The spreadsheet route breaks on three things, and they arrive in this order:
- Selectors rot.
span.a-size-medium.a-color-base.a-text-normalis an Amazon build detail. When it changes, your scenario silently writes empty names. - You cannot tell a paid placement from an earned one. Every card in
div.s-result-item.s-asinlooks identical to a CSS selector, so a sponsored slot and a rank 3 organic result land in the same Airtable column. - One page is 23 products. Pagination in a visual scenario means duplicating modules.
Lane two fixes all three, because ScrapingBee parses Amazon server side and hands back the labels.
Three endpoints, no selectors, no HTML:
| Endpoint | Required parameter | Returns |
|---|---|---|
/api/v1/amazon/search |
query |
products, products_count, refinements, url, page |
/api/v1/amazon/product |
query (a 10 character ASIN) |
56 fields including price, brand, rating, product_details, variations |
/api/v1/amazon/pricing |
asin |
pricing array, one entry per seller offer |
Authentication is a header. Send Authorization: Bearer YOUR_API_KEY on every request. The api_key query parameter still answers but the documentation now marks it deprecated, so build against the header.
The pricing endpoint takes asin, not query. Search and product both take query. Pricing does not, and it rejects it with a precise error rather than a guess:
{"errors":{"query":{"asin":["Missing data for required field."],"query":["Unknown field."]}}}A rejected request like that one is billed 0 credits, confirmed by a spb-cost: 0 response header on the failed call.
import requests
r = requests.get(
"https://app.scrapingbee.com/api/v1/amazon/search",
headers={"Authorization": "Bearer YOUR_API_KEY"},
params={"query": "fitness tracker"},
)
data = r.json()
print(data["products_count"], "products")One page of fitness tracker returned 23 products and a refinements object carrying every Amazon facet for that query, including brands, band_color, battery_average_life and case_diameter. Cost was 5 credits, read off the spb-cost header.
This is the part CSS selectors cannot give you. Each item in products carries its own placement labels:
| Field | Type | What it tells you |
|---|---|---|
is_sponsored |
boolean | Paid placement |
sponsored_position |
integer or null | Rank within the paid slots |
organic_position |
integer or null | Rank within the earned results |
is_amazons_choice |
boolean | Amazon's Choice badge |
best_seller |
boolean | Best Seller badge |
sales_volume |
string | Text such as 10K+ bought in past month |
So separating paid from earned is a filter, not a parsing problem:
featured = [p for p in data["products"] if p["is_sponsored"]]
organic = [p for p in data["products"] if not p["is_sponsored"]]
badged = [p for p in data["products"] if p["is_amazons_choice"] or p["best_seller"]]Across the live fitness tracker page, is_sponsored, is_amazons_choice and best_seller all came back with a mix of true and false values, so the labels are populated rather than stubbed. is_prime was false on every row of that page, which is worth knowing before you build a Prime filter on top of it.
Sorting is server side too. sort_by accepts featured, most_recent, price_low_to_high, price_high_to_low, average_review and bestsellers. The Amazon featured products API and Amazon search API pages cover the same surface from the product side.
const axios = require('axios');
const res = await axios.get('https://app.scrapingbee.com/api/v1/amazon/product', {
headers: { Authorization: 'Bearer YOUR_API_KEY' },
params: { query: 'B0GTMTZF3V' },
});
const p = res.data;
console.log(p.brand, p.price, p.currency, p.rating, p.reviews_count);That ASIN was taken from the live search response above rather than invented, and the call returned brand: "Fitbit", price: 99.99, currency: "USD", rating: 4.3, reviews_count: 2036. The full response carries 56 top level keys, including bullet_points, category as a breadcrumb ladder, featured_merchant with the seller id, rating_stars_distribution, sales_rank, variations and technical_details.
curl "https://app.scrapingbee.com/api/v1/amazon/pricing?asin=B0GTMTZF3V" \
-H "Authorization: Bearer YOUR_API_KEY"Returns a pricing array. Each offer has seller, seller_id, condition, price, price_shipping, currency, rating_count and a delivery_options list. On the ASIN above that was a single Amazon.com offer at 99.99 USD with free delivery.
If you want lane two without writing the HTTP layer:
pip install nocode-amazon
npm install nocode-amazonfrom nocode_amazon import AmazonScraper
bee = AmazonScraper("YOUR_API_KEY")
page = bee.search("fitness tracker")
for item in bee.featured(page):
print(item["sponsored_position"], item["title"])Measured, not quoted:
| Call | Credits |
|---|---|
Amazon search, default light_request=true |
5 per page |
Amazon search, light_request=false |
15 per page |
| Amazon product, default | 5 |
Amazon product, light_request=false |
15 |
| Amazon pricing | 5 |
| Any screenshot request | 15 |
| Rejected request (validation error) | 0 |
Light requests skip the browser. They are the default and they were enough for search, product and pricing in every call made here. Turn them off when you need review text or other content that only appears after JavaScript runs. Plan tiers are on the pricing page.
Both search and product accept country, currency, device, domain, language, zip_code, add_html, screenshot and tag. search adds pages, category_id, merchant_id, sort_by and autoselect_variant. domain is how you move marketplaces: co.uk, de, in and so on. zip_code matters more than it looks, because Amazon prices and delivery promises are postal code dependent.
Failed requests are retried internally for up to 30 seconds, so set client timeouts above that or you will abandon calls the API is still working on.
Public Amazon listing and product pages only. Nothing here signs in, and scraping under login credentials is prohibited by ScrapingBee's terms of service. Amazon's own Conditions of Use govern what you may do with the data once you have it, and the Airtable API and Make documentation cover the two tools in lane one.
Full parameter reference: Amazon API documentation and data extraction rules.
Can I scrape Amazon without writing code? Yes. The Make scenario above is the whole build, and the extraction rules are JSON rather than a program. You still need a ScrapingBee API key and an Airtable base.
How do I get only the featured products from an Amazon search?
Filter the products array on is_sponsored. The endpoint labels every result, so no parsing is involved. sort_by=featured changes the ordering Amazon returns.
Why is my price field empty in the no code lane?
span.a-offscreen moved, or the card you matched has no price because it is an ad unit. This is the failure mode that pushes people to lane two, where prices are parsed server side.
Which Amazon marketplaces work?
Any of them, through domain. Pass the top level domain, for example domain=co.uk or domain=in.
MIT. See LICENSE.