ScrapingBee Alternatives (Flat Price, No Credits)
Looking for a ScrapingBee alternative without per-request credits? The real credit math, five flat-price options, and a two-minute test before you switch.
On this page
ScrapingBee’s entry plan costs $49 a month for 250,000 credits; a stealth-proxy request burns 75, so the same plan covers 250,000 easy pages or 3,333 protected ones. That spread is why people hunt for a ScrapingBee alternative, and why most comparison pages disappoint: they rank one credit-metered API against another. This page does the opposite: credit math from ScrapingBee’s own docs, five alternatives with flat costs, and a two-minute test for whether you need the unblocking hardware at all.
The short answer
| Tool | What it costs | JS rendering | Unblocking | The catch |
|---|---|---|---|---|
| Scrapy + Playwright | $0, your machine | Via scrapy-playwright | None, you own it | You write and maintain every spider |
| Crawlee | $0, your machine | Built-in browser pool | None, you own it | Node or Python runtime to babysit |
| crawl4ai | $0, your machine | Yes, Playwright-based | None | Config surface; tuned for LLM input |
| Magpie Desktop | $49/mo flat licence, runs cost 1 not credits | Yes, automatic | Detects bot challenges, never solves them | Local only, one machine, no hosted API |
| Screaming Frog | £199/yr per licence | Yes | None | Exports are SEO-audit-shaped, not data feeds |
Ranked by how truly flat the cost is: the three free tools cost nothing but your time, Magpie is a fixed monthly licence, Screaming Frog is a fixed annual one. Every hosted API, including the ones calling themselves ScrapingBee killers, sits below all five on that axis. More on those at the end.
What ScrapingBee pricing actually buys
ScrapingBee’s help center publishes the exact credit cost per request type. The multipliers are the product:
| Request type | Credits |
|---|---|
| Classic proxy, no JavaScript | 1 |
| Classic proxy + JavaScript rendering | 5 |
| Premium proxies, no JavaScript | 10 |
| Premium proxies + JavaScript rendering | 25 |
| Stealth proxies | 75 |
| Any Google URL | 15 |
| Agentic Search API | 3,750 |
Run the arithmetic on ScrapingBee credits and the meter gets honest fast. Start from the $49 Freelance plan and its 250,000-credit balance. Plain pages cost about 0.02 cents each. Pages that need premium proxies and rendering cost 0.49 cents, a 25x jump for the same plan. Stealth-proxy pages cost 1.47 cents each, which means the plan’s 250,000 credits shrink to roughly 3,300 pages a month. Google SERP pulls run 16,666 per month. The Agentic Search API? About 66 queries.
What you’re renting at those rates is real infrastructure: a rotating residential and stealth proxy fleet, CAPTCHA handling, a Chrome-rendering farm that runs your requests in parallel. All of it has genuine per-request cost, which is why every company that builds it meters you somehow. Flat-price tools avoid the meter by giving you none of it.
A confession from researching this piece: pointing my local scraper at ScrapingBee’s own pricing page returned the same zero-word refusal my G2 test did. Their product exists for a reason.
The two-minute test before you pay anyone
Before choosing any tool, find out whether your actual target sites fight back. Not the worst-case site you imagine, the ones in your task list. I ran that test today with Magpie’s CLI on two pages of each kind.
First, a page that needs JavaScript but fights nobody. The DataTables ajax demo renders its table client-side, so plain HTTP gets an empty shell:
$ magpie scrape https://datatables.net/examples/data_sources/ajax \
--render browser --page-format markdown
# Ajax sourced data
...
| Airi Satou | Accountant | Tokyo | 5407 | 2008-11-28 | $162,700 |
| Angelica Ramos | Chief Executive Officer (CEO) | London | 5797 | 2009-10-09 | $1,200,000 |6.7 seconds wall-clock, 324 words of clean markdown, a proper table. Cost: zero, because the rendering happened on my machine. (Abridged output; full markdown includes the page’s explanatory text.)
Second, a page that does fight back. G2 review pages sit behind bot protection, and pointing the same tool at one returns exactly nothing:
Error: clean: extract: text and comments are not long enough: 0 0Zero words, zero records. That refusal is the point. Local tools hit the wall and stop; they don’t rotate in a residential proxy and retry in disguise. If your targets behave like the first page, you never needed the meter. If they behave like the second, no flat-price tool on this page will save you, and you should shop for the meter with the cheapest multipliers for your request shape instead of shopping for “no credits” as a fantasy.
For the blocking case in depth, I keep a separate guide on scraping without getting blocked. For the JavaScript case, see scraping dynamic websites.
The five alternatives
Scrapy plus Playwright: the free one you maintain
Scrapy is the BSD-licensed Python framework that has outlived every prediction of its death, current in the 2.x line with tens of thousands of GitHub stars. It handles scheduling, retries, deduplication, and export pipelines. Pair it with scrapy-playwright and it renders JavaScript pages too. Total software cost: $0.
The price is labor. Every site gets a spider, every spider gets maintained, and when the site redesigns, you get a support ticket from your own pipeline. Proxy rotation is a separate bill with a separate vendor, priced per gigabyte of residential traffic, and CAPTCHAs are entirely your problem. I’d pick this when scraping is part of your engineering job anyway and the targets don’t run hostile defenses. If you’re weighing it against other frameworks, the crawl4ai alternative page covers the LLM-shaped option in the same slot.
Crawlee: the free one from a metered company
Crawlee is Apache-2.0, built by the Apify team, available in TypeScript and Python. It gives you browser pools, request queues, proxy rotation helpers, and error handling that were hardened on Apify’s own platform. As free software, it is the most production-shaped of the three open-source options.
Two honest catches. First, it’s a framework, so you still write the crawler, manage the Chromium binary, and own the output format. Second, the parent company’s business is metered compute: Apify charges compute units, and it’s instructive that Crawlee’s own makers didn’t find a way to make hosted scraping flat. Use Crawlee when you want library-quality ergonomics at $0 and you’re already in Node. The Apify alternatives page has the full breakdown of what the hosted sibling costs.
crawl4ai: the free one that speaks LLM
crawl4ai is Apache-2.0 licensed and built around one output goal: clean markdown and structured JSON that a language model can eat without a token-burning preprocessing step. It runs Playwright under the hood, supports schema-driven extraction, and has grown fast enough to be the default answer in LLM-scraping threads.
The cons are the flip side of the hype. The configuration surface is wide, quality of the “fit markdown” extraction varies by site, and you’re running a Playwright runtime that needs its own update discipline. Unblocking: none. It’s the right pick when the consumer of your scrape is a model and the sites are polite.
Magpie Desktop: the flat licence, no credits, with the limits stated
Magpie is the tool I use for this job, so discount me accordingly, but the pricing model is the relevant fact here. The core CLI is open source under AGPL-3.0-or-later (or a commercial licence) and has no run counter at all. The desktop app is free to download with 5 free runs per install, then a flat $49 a month, one computer. A crawl costs one run whether it covers 50 pages or 50,000. A watch counts once. Nothing in the pricing moves with page count.
What the flat fee buys, verified in code and docs rather than marketing: pages escalate from plain HTTP to a real browser when JavaScript is required. Login sessions persist per site. Schema-driven extraction runs through your own LLM key, local Ollama included. And the selector cache re-synthesizes selectors a site redesign breaks. It respects robots.txt and throttles itself per host.
Now the limits, plainly, because this page exists to be honest about them: Magpie has no hosted API, so there’s no endpoint to call from a serverless function, and the data only ever moves on your machine. It detects bot challenges and reports them; it does not solve CAPTCHAs or impersonate its way past aggressive protection, by design. If your targets are walled, it fails exactly like the G2 test above. If they’re not, the $49 is the whole bill. The dynamic-websites guide shows the rendering behavior on real pages.
Screaming Frog: the flat licence that’s been flat for a decade
Screaming Frog’s SEO Spider is a desktop crawler, free up to 500 URLs per crawl and £199 a year per licence beyond that. It renders JavaScript, spiders at sizes most APIs would charge hundreds of dollars for, and has been the SEO industry’s default since before credit-based scraping APIs existed.
It’s on this list because the pricing is flat, and off the recommendation list for most readers because its output model is SEO-shaped: crawl analysis, response codes, titles, meta, audit views. Element extraction exists, but building a recurring data pipeline on it means fighting the tool’s center of gravity. Right tool if your scraping is actually a site audit. The Screaming Frog alternatives page goes deeper.
Why most ScrapingBee alternatives still meter you
Most pages ranking for this query pitch other hosted APIs as the escape without mentioning that all of them bill per request. The differences are real but they’re meter-shape differences, not flat-versus-metered. Zenrows starts at $16 a month (annual billing) for 45,000 credits and uses the same multiplier structure ScrapingBee does: JavaScript rendering costs 5 credits, premium proxies 10, both together 25. ScraperAPI hands out 1,000 free credits with 5 concurrent connections and meters the same way. Firecrawl’s Hobby tier runs $19 a month for 5,000 credits, and its per-page credits only look cheap until a crawl multiplies; the Firecrawl alternatives page has that trade-off in detail. I compared ScrapingBee and ScraperAPI feature by feature on the ScrapingBee vs ScraperAPI page if you want that level of detail.
There’s no scam here. For protected targets at volume, these companies deliver something genuinely hard, and their bills are honest about it. Just don’t confuse switching from one meter to another with owning your scraping.
Pick by your target sites
- Pages load fine with plain HTTP from your machine: any free tool above works; pick by language.
- Pages need JavaScript but don’t block you: still local territory. Magpie escalates automatically, or you wire Playwright yourself.
- Pages sit behind aggressive bot protection and you need thousands per month: that’s the meter’s job. Compare multipliers against your real request shape.
- One-off audits or small jobs: Screaming Frog’s free 500-URL tier or Magpie’s 5 free runs cover it without a card.
FAQ
Is there a scraping API with no credits at all?
A few services bill flat per request instead of per credit, but that’s still a per-request meter with a different unit. Truly flat means running the software yourself, which is what the free tools and Magpie do.
How much does ScrapingBee really cost per page?
On the $49 Freelance plan: about 0.02 cents for plain pages, 0.49 cents with premium proxies and rendering, and about 1.47 cents for stealth-proxy pages, all from ScrapingBee’s published credit table.
Is Magpie a full ScrapingBee replacement?
For sites that don’t block your machine, yes, with flat pricing and local data. It is not a replacement for unblocking: no residential-proxy rotation, no CAPTCHA solving, no hosted endpoint.
Can I self-host an unblocking API?
Partly. Proxy rotation and headless rendering you can self-host, and libraries like Crawlee help. CAPTCHA-defeating infrastructure is a paid arms race you probably don’t want to run.
What’s the cheapest way to scrape 100,000 pages a month?
If the pages aren’t walled, your own machine costs nothing, or $49 flat with Magpie. If they are walled, no flat tool helps; compare per-request rates and pick the cheapest multiplier for your mix.
Run the test tonight
Point any scraper at three of your real target pages, no proxies, no special headers. Count how many come back with real content. My bet from running this test on dozens of sites: most of them. Whatever number you get, now you know whether you’re shopping for a flat tool or a cheaper meter, and the tables above are the shortlist for each.
