Per credit, per response, per record or per compute unit: the four ways scraping APIs charge

Credits on a monthly plan

The common model: a monthly fee buys a pool of credits, a plain request spends one, a harder request spends more. ScrapingBee, ScraperAPI, ZenRows, Scrapfly, Scrapingdog, Scrape.do and ScrapingAnt sell this way, with entry plans between $16 and $49 a month. The pool resets monthly at most of them; ZenRows prints rollover per tier and ScraperAPI's FAQ says credits do not roll over. The comparison to make is the plan price divided by credits, then by the multiplier your targets need, which the guide on credits walks through.

Per successful response, no plan

Zyte sells no plan sizes: a rate per 1,000 successful responses that depends on which tier it assigns the target site and whether the response is rendered, $0.13 to $1.27 per 1,000 over HTTP pay as you go and $1.01 to $16.08 rendered, lower with a $100 or $200 monthly commitment. Crawlbase does the same by volume instead of by site: one graduated ladder from $3.00 per 1,000 for the first thousand requests down the brackets, with optional monthly blocks at a lower effective rate. Decodo sits between: plans from free to $99, but every mode priced per 1,000 requests on the plan.

Per record

Bright Data's Web Scraper API charges per record, one extracted item shown as a row, for instance one LinkedIn profile: $1.50 per 1,000 pay as you go, a plan with 384,000 records included and $1.30 per 1,000 beyond it, 5,000 records a month free. A record can take several requests to produce, so the per-record price is not comparable with a per-request one; it is comparable with the time it would take to build the scraper for that site yourself.

Per compute unit

Apify charges for the compute an Actor uses, $0.20 a unit on its free and $19 plans, $0.16 on $199 and $0.13 on $999, with each plan's fee credited back as usage. Many of the 70,495 Actors in its store add a price per result set by their author. There is no per-page figure to print; the cost is per Actor, on its store page. Firecrawl is the fourth shape's cousin: one credit per page, but the page is the cleaned text for a model, and the plans are small, 5,000 pages for $16 a month at annual billing.

Free tiers, read carefully

Monthly allowances: ScrapingAnt 10,000 credits, ZenRows 5,000, Bright Data 5,000 records, Scrape.do 1,000 successful calls with every feature, Firecrawl 1,000 credits, Apify $5 of usage. One-time allowances: Crawlbase 5,000 requests, ScrapingBee 1,000 credits, Scrapfly 1,000 credits with no time limit, ScraperAPI 1,000 credits with five concurrent connections, Scrapingdog 200 credits. Zyte gives $5 of credit for thirty days. Decodo's free plan has no credit count on the page; it is the $19 plan's rates at 10 requests a second. Oxylabs offers a trial whose terms were not on the pages read.

• Monthly or once: the word on the plan card decides whether the free tier is a test or a budget
• Card or no card: every vendor above except Oxylabs and Zyte states no card on the pages read; Zyte's $5 credit does not state either way
• Every feature or a subset: Scrape.do and ZenRows say everything is unlocked; ScraperAPI's free plan caps concurrency at five