Scraping APIs, and what a credit actually costs
A scraping API sells you the finished page rather than the address you fetch it from. That changes what you are buying and it changes the unit on the bill: 7 of the 8 we track charge in credits, and a credit is not a fixed amount of work. On one plan, a page that needs a browser costs more than a plain fetch.
So the per-gigabyte figure this site compares proxies on does not reach this product, and we do not invent one. What follows is what each vendor states, in its own unit, with the date we read it.
What one unit of work costs
The number that decides a bill on a per-request product, and the one nobody tabulates. These are the vendors’ own figures, per request type, not converted to a common unit — converting them would need an assumption about which tier you buy, and that assumption would be ours rather than theirs.
- 2 creditsScreenshot capture
- 1 creditMonitor
- 2 creditsSearch
- 2 creditsInteract
- 1 creditStandard page
- 1 creditSimple http scrape
- 1 creditBasic request
- 5 creditsJavascript rendering
- 25,000 creditsResidential bandwidth
Read those as a rate card rather than a price list. ZenRows draws 5 credits for JavaScript rendering and 25,000 for a gigabyte of residential bandwidth from the same balance; ScraperAPI charges 1 for a standard page. Two providers quoting “1 credit” are not quoting the same thing until you know what a credit costs on the plan you would buy.
Who charges for failures
On a per-request product this is the single largest swing in what you actually pay, and 3 of the 6 we publish state it. A blocked request that still costs a credit turns an 80% success rate into a 25% premium on every page you keep.
- Crawlbase — states it bills only on success
- Firecrawl — does not say either way on the pages we read
- ScraperAPI — does not say either way on the pages we read
- Scrapfly — states it bills only on success
- ScrapingBee — does not say either way on the pages we read
- ZenRows — states it bills only on success
Not stated is not the same as charges for failures. It means you cannot tell from the page you are buying on, and the number you are comparing is missing a term. Ask before you buy, and keep the reply.
Whose network is underneath
A scraping API is a storefront over somebody’s proxy pool. Of the 6 we publish, none states whose.
We are recording the gap rather than waiting for it to close, because the question is not academic. NetNut’s network was resold and white-labelled, and when the FBI seized it, every storefront in front of it inherited the problem. A clean-looking API may resell a pool that has since been seized as a botnet, and where the addresses come from does not stop at the vendor boundary.
Every scraping API we track
- Crawlbase — bills in credit, offers JavaScript rendering
- Firecrawl — bills in credit, 2 concurrent requests
- ScraperAPI — bills in credit, 20 concurrent requests, offers JavaScript rendering
- Scrapfly — bills in credit, offers JavaScript rendering
- ScrapingBee — bills in credit, offers JavaScript rendering
- ZenRows — bills in credit, 5 concurrent requests, offers JavaScript rendering
Listed, not yet reviewed
Our collector has read these and nobody has reviewed them, so they carry what it established and blanks where it did not. They have no page, and the blanks are the honest part: Serpapi (bills in credit), Zyte (bills in request).
Is a scraping API cheaper than proxies?
It depends on page weight, and the crossover is sharper than most people expect: a per-request price wins on heavy pages and loses on light ones, because a proxy bill is a function of bytes and an API bill is a function of pages. We worked the break-even out, and what your pages actually weigh is the input that decides it.
Questions
How do scraping APIs charge?
7 of the 8 scraping APIs we track bill in credits rather than by the gigabyte. A credit is not a fixed amount of work: on the same plan a page that needs a browser costs more than a plain fetch, so a per-gigabyte comparison does not reach this product.
Do scraping APIs charge for failed requests?
3 of the 6 we publish state that they bill only on success. The rest do not say either way on the pages we read, which is not the same as charging for failures. It means you cannot tell from the page you are buying on.
Whose proxy network does a scraping API use?
Of the 6 scraping APIs we publish, none states whose network is underneath. It matters because a storefront can resell a pool that is later seized, as happened to every reseller of NetNut's network.
Prices on this page were read from each provider’s own site. The oldest was checked . How we check