Firecrawl vs Crawl4AI: Hosted API or Open-Source Crawler?

2. What you are actually buying

Firecrawl’s official docs describe a context API for AI agents: search, scrape, and interact with the live web, then return Markdown or structured data. Documented capabilities include scrape, search, interact, map, crawl, parse, an agent, browser sessions, webhooks, and MCP. This comparison prices the public hosted plans (docs also mention a self-host path).

Crawl4AI’s official docs describe an Apache-2.0 crawler that turns sites into LLM-ready Markdown. Self-host with pip install -U crawl4ai, crawl4ai-setup, then AsyncWebCrawler — or Docker. The same docs and site also sell Crawl4AI Cloud (/scrape, /extract, /search, /serp, /answer, MCP). Their own table leaves browsers and proxies with you on self-host, and lists web search as a cloud capability.

So the real fork is Firecrawl hosted versus Crawl4AI self-host. Firecrawl hosted versus Crawl4AI Cloud is a closer comparison, but the credits are still not the same unit, and Crawl4AI’s cloud table is marked as a soft launch.

Checked November 6, 2026 on Firecrawl’s official pricing page and billing docs, and on Crawl4AI’s official docs, site, GitHub positioning, and live /v1/prices table. Do not convert credits 1:1.

Firecrawl hosted plans

Hosted figures from Firecrawl’s official pricing page, November 6, 2026 (USD). Annual billing: Hobby $190/year, Standard $990/year, Growth $3,990/year, Scale $7,190/year.

Official rate limits, Free through Scale: /scrape, /map and /search at 10 / 100 / 500 / 5,000 / 10,000 requests a minute; /crawl and /agent at 2 / 20 / 100 / 1,000 / 2,000; /interact at 2 / 20 / 100 / 1,000 / 1,500.

Credit spend, from the official pricing page and billing docs:

• Scrape or crawl: 1 credit per page. Billing docs list PDF parsing as +1 credit per PDF page.
• Map: 1 credit per call. Search: 2 credits per 10 results, rounded up (11 results = 4 credits). Scraping those hits adds the scrape cost.
• JSON, question/query, and highlights: +4 credits per page each (a JSON scrape is 5 credits). Prompt-injection check, PII redaction, audio, and video also add 4 each and stack (JSON + PII = 9).
• Interact: pricing page lists 2 credits per browser minute. Billing docs: one-minute minimum; prompt sessions bill at 7 per minute; Playwright code-only at 2; zero-data-retention adds 2 on top.
• Agent: 5 free daily runs, then dynamic pricing. Enhanced proxy: same 1 credit as a basic request; an escalated retry is not charged separately.

If Firecrawl returns a document, you pay — including a captured 403 or 404. If it returns no document, a basic scrape is not charged, with documented exceptions (a prompt-injection check that already ran, or interact time already consumed). Unused plan credits do not roll over on Free, Hobby, Standard, or Growth. The pricing table lists Scale rollover as one month; billing docs specify that for annual Scale (one month) and annual Enterprise (two months).

Pay-as-you-go on paid plans adds credits in $5 increments: 1,000 on Hobby, 2,000 on Standard, 2,500 on Growth, 5,000 on Scale. Bought credits stay until used and expire if you cancel. Free has no pay-as-you-go.

Crawl4AI official facts

Official GitHub and docs describe Crawl4AI as Apache-2.0: self-host, or use the hosted cloud when you do not want to run browsers and proxies. Documented self-host capabilities include Markdown for RAG, CSS / XPath / LLM extraction, and browser control.

Cloud docs and site: 1,000 free credits every month, no card, 10 requests a minute and 2 at a time. Paid throughput starts at 120 a minute and 10 at a time from $10 a month:

Official site credit hint: 1 credit = $0.001; a page at 0.5; search at 1; archive at 0.25; a real browser at 1; a hard site at 2. The live /v1/prices table matches (base scrape/extract 0.50; search/serp/answer 1.00) plus engine multipliers (archive 0.5×, light/standard 1×, heavy browser 2×, hard site 4×). Soft-launch warning: prices and the free credit can change; purchased credits never expire. Citation, not a CTA.

3. Real cost by scenario

Which published plan covers the work you actually run? Calculations use official rates checked November 6, 2026, for the workloads described — no retries, no JSON modifiers unless stated, no extra Crawl4AI residential-proxy multiplier. Hard, authenticated, or PDF-heavy pages move the number.

Scenario A — Prototype: 800 Markdown pages a month

Firecrawl basic scrape is 1 credit per page, so 800 credits fit the Free plan’s 1,000. Cost: $0, with 2 concurrent browsers and a 10/min scrape cap. The same 800 pages with JSON extraction cost 5 credits each (billing docs): 4,000 credits, which needs Hobby at $19/month (or $16/month billed annually).

Crawl4AI self-host: software documented as free; you pay your own CPU, browser, and proxies (UA is not inventing that invoice). Crawl4AI Cloud at 0.5 credit per page is 400 credits, inside the free 1,000. If every page bills as a hard site at 2 credits, 800 pages become 1,600 credits, so Starter at $10/month is the next published step.

Scenario B — Production Markdown: 10,000 pages a month

Firecrawl needs 10,000 credits. Hobby includes only 5,000. Two official paths: stay on Hobby at $19 and buy 5,000 extra credits (1,000 credits per $5, so $25) for a total of $44/month; or jump to Standard at $99 for 100,000 credits and leave 90,000 unused that will not roll over. Hobby’s 100 scrape requests a minute and 5 concurrent browsers bind you only if you burst the job. Pay $99 when you need Standard’s 25 browsers, not just its credit pile.

Crawl4AI Cloud at 0.5 credit per page is 5,000 credits, inside Starter’s 10,000, at $10/month. The Crawl4AI site itself publishes that comparison ($10 buys 20,000 pages vs Firecrawl Hobby’s 5,000, using Firecrawl prices dated October 14, 2026). If all 10,000 pages bill as hard sites at 2 credits, you need 20,000 credits: Starter plus 10,000 extras at $1 per 1,000, for $20/month — or Builder at $99.

Scenario C — 50,000 Markdown pages a month

Firecrawl wants 50,000 credits. Hobby plus pay-as-you-go would mean 45,000 extra credits ($225) plus $19, for $244. Standard includes 100,000 credits at $99 — so Standard wins here. You still have 50,000 unused credits that will not roll over. Growth at $399 is not required by credits alone.

Crawl4AI Cloud at 0.5 credit per page is 25,000 credits. Starter covers 10,000; 15,000 extras at $1 per 1,000 add $15, for $25/month. If every page is a hard site at 2 credits, you need 100,000 credits: Builder at $99, or Starter plus $90 of extras — roughly a wash before throughput (Builder 600/min and 20 concurrent vs Starter 120/min and 10).

Scenario D — 10,000 pages with JSON extraction

JSON format adds 4 credits, so 10,000 Firecrawl pages cost 50,000 credits: Standard at $99/month, or a painful Hobby top-up ($225 + $19 = $244). Add PII redaction and the billing docs’ stacking example takes each page from 5 credits to 9; 10,000 pages become 90,000 credits — still inside Standard’s 100,000, but barely.

Crawl4AI’s live price table lists extract at the same 0.50 base as scrape. That does not mean 0.50 buys the same LLM-schema extraction Firecrawl bills at 5 credits. Price each vendor’s actual extract path against a sample of your pages before you assume a 10× gap.

A prompted Firecrawl interact session bills at 7 credits per browser minute (one-minute minimum). Three hundred prompted minutes are 2,100 credits by themselves. A search-then-scrape loop is cheaper than it looks: 200 Firecrawl searches of 10 results cost 400 credits; add 1,000 Markdown scrapes and the month is 1,400 credits, inside Hobby at $19.

4. Technical workflow: how a buyer actually runs each one

Firecrawl hosted, as the official docs describe it: get an API key (Free plan, no card) and send a scrape call with a URL and a format. Docs also show Python and Node SDKs, a CLI, and MCP setup. You do not provision a browser. You do pick a format, because the format is the bill.

Documented production sequence: map a domain when you lack the URL list (1 credit per call); crawl or batch-scrape only the URLs you need, with an explicit limit (credits bill as pages complete); scrape one-off URLs, keeping Markdown as the default and using JSON, question, or highlights only where you need an LLM extract; search when the starting point is a query, remembering the round-up rule; interact only when a static scrape cannot finish the job, then stop the session. Watch metadata.statusCode — a captured 403 or 404 still costs a credit.

Concurrency is part of the workflow. Free: 2 concurrent browsers; Hobby: 5; Standard: 25. Extra jobs wait in the published queue. If your runtime fans out 20 scrapes at once, Hobby will serialize most of them even when credits remain.

Crawl4AI self-host: install the package, run crawl4ai-setup once, call AsyncWebCrawler, and read result.markdown. The docs then cover Docker, hooks, proxies, stealth, session reuse, and CSS / XPath / LLM extraction. The hidden work sits around that snippet: keep the browser current, scale headless Chrome, bring proxies if targets block you, store output, handle retries, stay inside legal and contractual limits, and rebuild search / SERP / answer yourself (the docs list those as cloud endpoints). If that list is your normal Tuesday, self-hosting is the point of Crawl4AI. If it sounds like a second product you did not plan to staff, Firecrawl matches the job.

Crawl4AI Cloud: verify an email, receive 1,000 free credits each month, and reuse one bearer key for scrape, extract, search, SERP, answer, batch jobs, and MCP. Operationally closer to Firecrawl than to pip install. Still a different meter, still in soft launch, still not the hosted pick here.

5. Pitfalls that change the bill — or the project

1. Credits are not interchangeable. Firecrawl’s basic page is 1 credit. Crawl4AI’s published page rate starts at 0.5 and climbs with engine level. “Twice as many pages per dollar” without an engine level is a screenshot, not a model.
2. LLM formats multiply Firecrawl’s cost by five, or more. Markdown at 1 credit vs JSON at 5 is the Hobby-vs-Standard fork in scenario D. Question, highlights, PII redaction, audio, and video each add another 4, and they stack.
3. Interact is not scrape. Prompted browser time at 7 credits a minute will outrun a page-based model. If the agent “clicks around until it finds the price,” price the minutes.
4. 403s are not free, and most plan credits expire each cycle. A returned error document costs 1 Firecrawl credit. Prefer Hobby plus official $5 top-ups until a month actually needs Standard’s 100,000 credits or 25 browsers. Bought pay-as-you-go credits persist until used and vanish if you cancel.
5. Some Firecrawl URLs are not 1-credit scrapes. Billing docs price typical X/Twitter requests through Grok at 30 credits (1 + 29) and list LinkedIn profile or company pages as provider-routed, from 5 to 40 credits.
6. Crawl4AI Cloud is in soft launch. Re-read the live price table before you rely on a number. Purchased credits are described as never expiring; the free grant and list prices can move.
7. Self-host is unmetered by the vendor, not free. You still pay machines, proxies, debugging, and the hour a site changes its markup. UA will not invent those costs.
8. Search and legal limits still apply. Firecrawl search is a billed hosted endpoint; Crawl4AI’s docs place /search, /serp, and /answer on the cloud side of their table. Cleaner Markdown does not grant rights you do not have.

6. Decision: what to do Monday morning

• If you want a hosted API and you do not want to operate crawlers — start with Firecrawl. Use the official Free plan (1,000 credits, no card), run a slice of your real workload, and watch which formats you call. Stay on Free or Hobby until a month needs 100,000 credits or 25 concurrent browsers.
• If the crawler must run on infrastructure you control — install Crawl4AI, run crawl4ai-setup, and prove it against your own targets. A hosted SKU does not replace that promise.
• If you are only here because Crawl4AI Cloud is $10 — model easy pages, hard pages, and extract separately, then compare that number to Firecrawl Hobby plus official top-ups, not to Firecrawl Standard. You are comparing two hosted services, not “open source versus paid.”
• If the job is raw HTML collection or an actor store — you are on the wrong fork. Use the internal comparisons linked above.

When the hosted API is the job, Firecrawl is the recommendation. Next: measure credit burn on the official Free tier before you pick Hobby or Standard.

7. FAQ

Did UA test Firecrawl against Crawl4AI?

No. This page compares official product claims, prices, credit rules, and the published Crawl4AI /v1/prices table. It does not report success rates, Markdown quality, or speed. UA did not create evaluation accounts for this article.

Why recommend Firecrawl if Crawl4AI Cloud starts at $10?

Because $10 for 10,000 cloud credits and $19 for 5,000 Firecrawl credits are not the same product or the same credit. Firecrawl is the hosted API pick for buyers who want that vendor’s documented feature set and billing schedule without operating crawl infrastructure. Crawl4AI remains the open-source self-host option. Crawl4AI Cloud is not this page’s top pick.

Will Hobby at $19 be enough?

On paper, yes for 5,000 basic Firecrawl pages — fewer once JSON, interact minutes, or X/LinkedIn provider routes enter the mix. Official top-ups are $5 per 1,000 credits, so a 10,000-page Markdown month can stay on Hobby for $44. Move to Standard when you consistently need 100,000 credits or 25 browsers.

Can these prices change?

Yes. Every figure here is a dated snapshot from official sources checked November 6, 2026, not a quote. Crawl4AI’s live table warns that soft-launch prices and the free credit can change. Recheck the named pages before you buy. More on this site: Firecrawl review, best scraping APIs for LLMs, Scraping hub.

8. Sources

Official pages consulted November 6, 2026. Named for verification, not linked.

• Firecrawl official pricing page (checked November 6, 2026)
• Firecrawl official billing docs (checked November 6, 2026)
• Firecrawl official documentation / introduction (checked November 6, 2026)
• Firecrawl official homepage (checked November 6, 2026)
• Crawl4AI official documentation (checked November 6, 2026)
• Crawl4AI official site, including the published cloud plans and credit hint (checked November 6, 2026)
• Crawl4AI live /v1/prices table (checked November 6, 2026; the file itself notes a soft launch)
• Crawl4AI official GitHub repository positioning (Apache-2.0, self-host plus cloud; checked November 6, 2026)