Firecrawl vs ScraperAPI: LLM Crawl or Scrape API?

2. What you are actually buying

Firecrawl hosted plans (official, November 6, 2026)

Free $0 / 1,000 credits / 2 concurrent browsers; Hobby $19 or $16 annual display / 5,000 / 5; Standard $99 or $83 ($990/year) / 100,000 / 25; Growth $399 or $333 ($3,990/year) / 500,000 / 50; Scale $749 or $599 ($7,190/year) / 1,000,000 / 100. Rate limits Free→Scale: scrape/map/search 10 / 100 / 500 / 5,000 / 10,000 per minute; crawl/agent 2 / 20 / 100 / 1,000 / 2,000; interact 2 / 20 / 100 / 1,000 / 1,500.

Credit reminders: scrape or crawl 1 per page; JSON +4 (JSON scrape = 5); search 2 per 10 results rounded up; map 1 per call; Interact 2 per minute on the pricing page versus 7 per prompt-minute in billing docs; Agent 5 free daily runs then dynamic; returned 403/404 documents are charged; no plan-credit rollover on Free through Growth.

ScraperAPI official plans (November 6, 2026)

Annual billing is 10% off the monthly prices below.

Official credit weights: standard page 1; Amazon 5; Google and Bing (including subdomains) 25; LinkedIn 30; Cloudflare, Datadome, or PerimeterX bypass +10. Credits do not roll over. Exceeding concurrency returns HTTP 429 without an extra credit charge. Running out of credits pauses the account; Hobby through Professional document auto-upgrade, and Scaling and above document pay-as-you-go. We do not add marketplace-specific weights (for example Walmart or eBay) that the official FAQ we fetched did not name.

3. Real cost by scenario

Same workload, two official price lists. No retries. Firecrawl figures use basic Markdown scrape = 1 credit unless JSON is stated. ScraperAPI figures use a standard page = 1 credit unless a named multiplier applies.

Scenario A — 8,000 generic Markdown or HTML pages a month

Firecrawl: 8,000 credits. That overflows Hobby (5,000) and fits Standard (100,000). Cost: Standard $99/month (or $83/month billed annually). As JSON: 40,000 credits, still Standard.

ScraperAPI: 8,000 standard credits fit Hobby’s 100,000 with a large surplus. Cost: Hobby $49/month (or $44.10 annual). If every URL needed the Cloudflare/Datadome/PerimeterX +10, each page would be 11 credits and 8,000 pages would be 88,000 credits — still Hobby. This is the scenario where ScraperAPI’s larger credit packs look cheaper, because the job is “fetch pages,” not “return LLM context.”

Scenario B — 2,000 Google result pages a month, plus 6,000 generic pages

Firecrawl does not publish a “Google = 25 credits” rule. Those 2,000 URLs bill as ordinary scrapes (2,000 credits) or as search+scrape if you use the search endpoint (search is 2 credits per 10 results, rounded up, plus scrape of the hits). Budget the search endpoint separately if that is how you collect them. The 6,000 generic pages add 6,000 credits. A Markdown mix around 8,000 credits is again Standard if you are already over Hobby.

ScraperAPI: 2,000 Google pages × 25 credits = 50,000 credits, plus 6,000 generic = 56,000 credits. That still fits Hobby’s 100,000. Cost: $49/month. If the 2,000 Google pages also needed the +10 anti-bot add-on, they would be 35 credits each (70,000) plus 6,000 = 76,000 — still Hobby. Cross Hobby once Google volume grows: 4,000 Google pages alone are 100,000 credits, the entire Hobby bucket, before any other URL.

Scenario C — 10,000 pages as structured fields for an LLM

Firecrawl JSON: 10,000 × 5 = 50,000 credits → Standard $99/month. This is the case Firecrawl is built for: you pay more credits per page and you do not stand up your own extractor.

ScraperAPI: 10,000 standard credits still sit on Hobby at $49/month if the pages are generic HTML. You then pay engineering — your parser, your Markdown conversion, your schema drift — which we will not invent as a dollar figure. If those 10,000 pages were LinkedIn URLs, ScraperAPI bills 30 credits each: 300,000 credits, which needs Business at $299/month (3,000,000 credits) because Startup’s 1,000,000 is not enough. The “$49 vs $99” headline reverses as soon as the URL class changes.

Scenario D — 400 LinkedIn URLs a week (about 1,720 a month)

Firecrawl billing docs list provider-routed LinkedIn fetches in a 5–40 credit range. At the low end, 1,720 × 5 = 8,600 credits (Standard at $99 if you have no Hobby headroom). At the high end, 1,720 × 40 = 68,800 credits, still Standard. We will not pretend we know where your URLs land inside that range; measure once on a small batch before you budget the month.

ScraperAPI: LinkedIn is a flat 30 credits on the official FAQ. 1,720 × 30 = 51,600 credits, which fits Hobby at $49/month. If the same URLs also needed the +10 bypass, they would be 40 credits each (68,800) — still Hobby. Volume is what breaks this: 4,000 LinkedIn URLs × 30 = 120,000 credits, just over Hobby, so Startup at $149/month.

Scenario E — Prototype week

Firecrawl Free: 1,000 credits, no card, no top-up. ScraperAPI: 7-day trial with 5,000 credits and no card, or the ongoing free plan with 1,000 credits and 5 threads. If you only need to kick the tires of HTML fetching, ScraperAPI’s trial is the larger short window. If you need to kick the tires of Markdown or JSON-for-agents, Firecrawl Free is the relevant test — 200 JSON pages already consume the entire 1,000-credit bucket (200 × 5).

4. Technical workflow

Documented flows only. UA ran neither API.

1. Write down the output you owe the rest of the stack. Markdown or JSON for an agent points at Firecrawl. HTML for your parser points at ScraperAPI. If you cannot name the output, you are not ready to compare prices.
2. Count URL classes, not just URL counts. On ScraperAPI, generic / Amazon / Google-Bing / LinkedIn / anti-bot bypass are different credit lines. On Firecrawl, Markdown / JSON / search / interact / social fetches are different credit lines.
3. Firecrawl path: API key, /scrape for a URL, /map or /crawl for a site, /search when you must discover URLs, JSON only when the schema is real. Respect crawl’s tighter per-minute caps.
4. ScraperAPI path: API key, send the URL, enable JS rendering when the page needs it (included, not a surprise add-on on the official plans), and watch geotargeting: Hobby and Startup are listed with US and EU, not the global flag that starts at Business.
5. Plan for empty and blocked responses. Firecrawl charges when it returns a document, including captured 403/404. ScraperAPI documents 429 with no extra credit charge when you exceed concurrency, and a pause (or auto-upgrade / PAYG depending on plan) when credits run out.
6. Match concurrency to the plan, not to your crawler’s optimism. Firecrawl Hobby publishes 5 concurrent browsers and 100 scrapes a minute; Standard publishes 25 and 500. ScraperAPI Hobby publishes 20 threads; Startup 50; Business 100. A job that is “cheap on credits” can still need the next tier because it wants to run wide.
7. Keep geotargeting and search separate in the design doc. ScraperAPI’s US/EU versus global split is a plan feature. Firecrawl’s search endpoint is a credit feature (2 per 10 results). Mixing “we need French Google results” with “we need Markdown chunks” without assigning each need to a vendor is how teams buy the wrong one twice. Write both needs down before you open either dashboard.

If you already run one of the two and you are adding the other for a second job, isolate the keys and the logs. The failure modes (Firecrawl HTTP 402 on Free, ScraperAPI pause or auto-upgrade when credits hit zero, 429 on thread caps) look similar in a dashboard and mean different billing events.

5. Pitfalls

• Treating 100,000 ScraperAPI credits as 100,000 Firecrawl pages, or the reverse. Google at 25x and JSON at 5x destroy both mental models.
• Buying Firecrawl to save money on HTML. Scenario A is cheaper on ScraperAPI’s Hobby pack if you truly only needed HTML.
• Buying ScraperAPI to “do RAG cheaper,” then spending a sprint on Markdown, boilerplate removal, and schema extraction that Firecrawl already bills as a +4.
• Forgetting geotargeting. ScraperAPI Hobby and Startup list US and EU. Global targeting is documented from Business upward.
• Ignoring Firecrawl’s Free 402 versus ScraperAPI’s trial. They are not the same on-ramp. One is a recurring 1,000-credit month with no top-up; the other includes a 5,000-credit 7-day trial.
• Inventing extra marketplace weights. We only cite Amazon, Google/Bing, LinkedIn, and the +10 bypass listed on the official FAQ we fetched on November 6, 2026.

6. Decision

Our hosted pick for LLM context remains Firecrawl. That is a category call, not a claim that Firecrawl is cheaper per HTML page — Scenario A shows it often is not. If HTML is the job, ScraperAPI is the tool we would evaluate first in this pair.

7. FAQ

Is Firecrawl cheaper than ScraperAPI?

Not as a general rule. Firecrawl’s paid entry is $19/month for 5,000 credits; ScraperAPI’s is $49/month for 100,000 credits. Which invoice wins depends on whether you bill Markdown/JSON pages or HTML pages with ScraperAPI’s multipliers.

Can ScraperAPI replace Firecrawl for RAG?

Only if you build the Markdown or structured-extraction layer yourself. Official ScraperAPI materials we checked do not sell that layer as the headline product.

Does ScraperAPI include JavaScript rendering?

Yes. The official pricing page lists JS rendering and premium proxies as included on all plans.

How many credits is a Google page on ScraperAPI?

25, including subdomains, on the official FAQ checked November 6, 2026. Bing is the same. A Cloudflare, Datadome, or PerimeterX bypass adds 10.

Do you have an affiliate link for ScraperAPI?

Not at the time this page was written. A hidden placeholder sits in the HTML for later; it is not visible and it does not point at a live vendor URL.

8. Sources

Vendor pages are named in text so this article does not send click-throughs to non-affiliate destinations. Dates are Europe/Paris verification dates.

• Firecrawl official pricing page and billing documentation (checked November 6, 2026)
• Firecrawl official documentation (checked November 6, 2026)
• ScraperAPI official pricing page (checked November 6, 2026)
• ScraperAPI official pricing FAQ on credits, concurrency, rollover, refunds, and plan behavior (checked November 6, 2026)