Skip to content
CrawlfolioSourced research on web data for AI
Menu

Spider bills the bytes it downloads; Firecrawl bills the pages it returns. On today’s list prices that makes Spider far cheaper for big crawls of light pages (its own estimate for 1M pages is $28-$160, against $749 a month on Firecrawl Scale), and Firecrawl the safer bill for heavy pages, JSON extraction and anything you have to budget to the dollar. Spider pricing, Firecrawl pricing, verified 2026-11-17.

Two of the five pages that rank for this comparison are Spider’s own, and the rest are thin or carry stale numbers. So below I turn “$1 per GB” into a price per page, find the page size where the two cross over, and cover what moves real bills: failed pages, expiry, crawl defaults, rate limits and licences.

Short on time?

Here's my take

Large crawls of light, mostly static pages (docs, blogs, 100K-1M pages)
Spider pay-as-you-go
Spider's own estimate is $28-$160 for 1M pages, against $599-$749 on Firecrawl Scale. Check your average page weight first: the gap closes as pages get heavier.
Small or irregular jobs (under about 10K pages a month)
Spider, from a $5 top-up
Credits never expire and there's no subscription, against $41-$44 a month on Firecrawl. Under 1,000 pages a month, Firecrawl Free costs nothing.
Structured JSON, screenshots, actions or search in one API
Firecrawl Standard, $99
100K pages a month, or 20K pages as JSON. More output formats per call and nine official SDKs.
Heavy JavaScript or media pages, or a fixed monthly bill
Firecrawl
1 credit per page whatever the page weighs, with a spend cap on extra credits. Spider doesn't publish how it meters Chrome-rendered pages.
Bot-protected targets
Firecrawl, on the evidence there is
Scrapeway's independent October 2026 table puts Firecrawl at 81% success. No third party has tested Spider, so trial both on your own URLs.
Embedding or modifying the crawler in a product you sell
Spider's spider-rs crate (MIT)
A permissive Rust crate, against Firecrawl's AGPL-3.0 server, whose network clause applies if you modify it and offer it to others.

Prices verified . Sponsors cannot buy a verdict or a position.

Both sit in my crawl APIs hub. Firecrawl’s full plan ladder and add-ons live in Firecrawl pricing; here I keep only the rows this comparison needs.

Spider and Firecrawl side by side #

Spider wins on price per light page, expiry and licence; Firecrawl wins on predictability, output formats and SDKs. Hosted services, USD list prices, verified 2026-11-17.

Spider Firecrawl
What you pay for Bytes downloaded ($1/GB) plus CPU time ($0.0001/min) Credits per page from a monthly plan (1 a page, +4 for JSON)
Entry price $5 minimum top-up, no subscription Hobby $19 a month ($16 billed yearly), 5,000 credits
Free 2,500 credits ($0.25) once, if signup passes checks 1,000 credits every month, no card
Do credits expire? Never Plan credits reset monthly; bought credits expire if you cancel
A failed page Billed: bandwidth plus compute, 0.01-credit minimum No document = 0; a returned 403/404 = 1 credit
Rate limit 10,000 requests a minute on every plan 10 to 10,000 scrapes a minute by plan
Concurrency Unlimited by default on pay-as-you-go 2 to 100 browsers by plan
Default crawl cap None (limit 0), depth 25 10,000 pages, with a 402 if credits can’t cover it
JavaScript rendering smart by default; billed as compute, no multiplier Included; 1 credit either way
Proxies Rotation included, 199 countries, bring your own for 50% off transfer Basic (26 countries) or enhanced (US, Netherlands), +0 credits
Output raw, markdown, commonmark, text, xml, bytes markdown, html, json, screenshot, links, summary, images, audio and more
Open-source licence spider-rs crate, MIT Core AGPL-3.0, SDKs MIT
Official SDKs CLI, JavaScript, Python, Rust (Go per Spider’s blog) Python, Node, Go, Java, Elixir, Rust, Ruby, .NET, PHP, CLI

Sources: Spider pricing, Spider FAQ, Spider API reference, Firecrawl pricing, billing, rate limits, crawl docs, Firecrawl repository.

How does each one charge you? #

Spider meters what it downloads and how long it computes; Firecrawl counts pages against a monthly allowance. That one difference explains almost every result further down.

Spider bills bytes and CPU time, with no subscription #

Spider’s pricing page lists $1 per GB of “pre-transformation content” plus $0.0001 per minute of CPU, “capped per request”. One credit is $0.0001, the minimum top-up is $5, and packs run $5, $25, $100, $500+ (5% bonus) and $2,000+ (12% bonus). Spider pricing, Spider guide, verified 2026-11-17.

“Pre-transformation” matters for RAG work. You pay for the raw HTML Spider fetches, not the smaller markdown it hands back.

Spider’s guide fills in the compute side. Compute is 1 credit a minute, prorated to the millisecond, so “A 12 second request is 0.2 credits”, and API compute stops at 30 credits (50 with a headless browser). Spider guide.

In dollars, that 12-second request costs $0.00002 in compute, and the ceiling is $0.003 a request, or $0.005 with a browser. The guide also says “Credits never expire”, and the FAQ says there’s “no monthly subscription”. Spider FAQ.

Spider also publishes “typical” costs per request: about $0.0001 for /scrape, $0.0002 for /links, $0.0003 for /crawl, $0.0006 for /screenshot and $0.001 for /search. Zero data retention costs 2.5x credits. Spider pricing.

One conflict to know about: the pricing page says “Minimum top-up: $5”, while the guide says you can “start with as little as $1” and Spider’s comparison post says “you can start with $1”. I trust the pricing page, since it’s the checkout-facing one. Spider guide, Spider’s comparison.

Firecrawl sells page credits on a monthly plan #

A basic scrape or crawled page is 1 credit, JSON extraction adds 4, a map call is 1, search is 2 per 10 results and a PDF is 1 per page. The enhanced proxy adds nothing, and zero data retention adds 1. Firecrawl pricing, billing docs, verified 2026-11-17.

Plan Monthly Billed yearly Credits a month Extra credits per $5 Concurrent browsers
Free $0 $0 1,000 None 2
Hobby $19 $16 5,000 1,000 5
Standard $99 $83 100,000 2,000 25
Growth $399 $333 500,000 2,500 50
Scale $749 $599 1,000,000 5,000 100

Plan credits don’t roll over on Free through Growth, and yearly plans still reset monthly. The billing docs say annual Scale rolls over one month and annual Enterprise two, and bought credits “expire if you cancel your subscription”. Billing docs.

You can cap monthly pay-as-you-go spend in $5 steps, with 0 turning extra credits off and a blank meaning no limit. That cap is the main reason Firecrawl is easier to budget. Billing docs.

Long-exposure light trails at night, white, orange and blue streaks curving away on the left and bands of red streaks on the right, against black
Spider meters the data that flows through, Firecrawl counts the pages that come out. Light static pages barely move Spider's meter and heavy ones push it up, while Firecrawl charges the same credit for a static page as for a JavaScript one. Photo: Tim Rüßmann / Unsplash · Unsplash License

So what do three real jobs cost? #

At 10K, 100K and 1M pages a month, Spider’s own estimates come in far below Firecrawl’s cheapest plan. Firecrawl figures are my arithmetic on its list prices; Spider figures are Spider’s per-100-page estimates scaled up, so they hold only for pages like the ones Spider assumed. Verified 2026-11-17.

Job a month Firecrawl, monthly / yearly How Spider, Spider’s estimate Where Spider’s figure comes from
10K static pages $44 / $41 Hobby + 5 packs of 1,000 $0.20-$1.00 HTTP mode, $0.002-$0.01 per 100 pages; guide says $0.80-$1.00
10K JavaScript pages $44 / $41 Same: 1 credit either way $0.35-$1.55 Chrome mode, $0.0035-$0.0155 per 100 pages
10K pages as JSON $99 / $83 Standard (50K credits) Not published No per-page extraction price on the pages I read
100K pages $99 / $83 Standard $2.80-$12.30 Smart mode; guide says $8-$15 for 100K mixed pages
1M pages $749 / $599 Scale $28-$155 Smart $28-$123, Chrome $35-$155; guide says $80-$160 for a 1M-page RAG pipeline

Sources: Firecrawl pricing, Spider pricing, Spider guide.

Here’s the Firecrawl arithmetic. 10K pages is Hobby’s $19 plus 5,000 more credits at 1,000 per $5, so $19 + $25 = $44 ($16 + $25 = $41 yearly). 10K JSON pages need 50,000 credits, which on Hobby would be $19 + 45 × $5 = $244, so Standard at $99 is the cheaper route.

For 1M pages, Growth plus packs would be $399 + 200 × $5 = $1,399, so Scale at $749 wins. Per 1,000 pages that’s $4.40 at 10K, $0.99 at 100K and $0.749 at 1M, on monthly billing.

Even at the top of each Spider range, the gap is wide: about a fifth of Firecrawl’s price at 1M pages and a few percent at 10K. The catch is in Spider’s own numbers. Its highest Chrome estimate, $0.0155 per 100 pages, is what about 155 KB of bandwidth a page would cost at $1/GB with no compute, so these estimates describe light pages.

I left Spider’s JSON row blank on purpose. Spider has an AI endpoint, but none of the pricing pages I read give a per-page price I could put next to Firecrawl’s 5 credits. For more on credit-based pricing across vendors, see scraping API credits.

When does Spider stop being cheaper? #

When your average page gets heavy. Spider bills bytes and Firecrawl bills pages, so the two cost the same when a page’s metered size, priced at $1/GB, equals Firecrawl’s price per page.

Where the two cost the same (bandwidth only)

Firecrawl plan, fully used Firecrawl per page Spider costs the same at
Hobby + packs, 10K pages ($44) $0.0044 about 4.4 MB
Standard monthly, 100K pages ($99) $0.00099 about 990 KB
Standard yearly ($83) $0.00083 about 830 KB
Scale monthly, 1M pages ($749) $0.000749 about 749 KB
Scale yearly ($599) $0.000599 about 599 KB

My arithmetic on Spider and Firecrawl list prices, using decimal units (1 GB = 1,000,000 KB). Below the size in the right-hand column, Spider is cheaper on bandwidth.

Compute pushes the line down a little. A 12-second request adds $0.00002, which is the same money as 20 KB of bandwidth, so the Scale break-even drops from about 749 KB to about 729 KB if every page takes that long.

The ceiling matters more than the typical case. At Spider’s cap of $0.005 a request with a headless browser, compute alone would be $5,000 for 1M browser pages; Spider’s own examples are far lower, but the cap is the number to keep in mind for slow, script-heavy sites. Spider guide.

Two illustrations, both assumptions rather than quotes. 1M pages at 100 KB is 100 GB, or $100 of bandwidth against Firecrawl Scale’s $749. 1M pages at 2 MB is 2,000 GB, or $2,000, and Firecrawl is the cheaper bill.

Whether a page “weighs” 100 KB or 2 MB depends on what Spider counts. In plain HTTP mode, that’s the HTML response. In Chrome mode, a browser also loads scripts, styles, fonts and images, and Spider doesn’t say whether those bytes are metered.

What neither vendor publishes

  • How Spider meters Chrome-rendered pages. HTML only, or every subresource the browser loads: not stated on the pricing page, guide or API reference. This decides the break-even for JavaScript-heavy sites.
  • An independent speed or success test of Spider. Scrapeway and the Proxyway test summarised by Zyte both leave it out.
  • Keyless limits. Spider says keyless runs are “rate limited” and Firecrawl caps keyless access per IP per day; neither gives a number. Spider API, Firecrawl rate limits.
  • Firecrawl’s enhanced proxy pool type. The proxies page doesn’t say whether enhanced means residential.

If your pages are mostly server-rendered HTML, the arithmetic favours Spider by a lot. If they’re media-heavy or need a browser, I’d ask Spider how it meters them before moving a big crawl.

What happens when a page fails? #

Spider bills it; Firecrawl usually doesn’t. Two of the comparisons I read get this wrong, most likely because Spider changed its terms.

Spider’s pricing FAQ says “A failed page is not free”: each attempt bills bandwidth and compute “whatever status the page returns”. The guide sets a 0.01-credit minimum for a blocked or empty page. Spider pricing, Spider FAQ, Spider guide.

Spider’s own guide still says “A failed API request costs nothing” in its section on model calls. I read that as covering AI model calls, not page fetches, and I trust the pricing FAQ because it’s current and specific to pages. Spider doesn’t say which reading is right.

Firecrawl’s rule: a request with no result “is not charged”, but a page that comes back with a 403 or 404 “costs 1 credit”. Threat scans (2 credits a URL) and prompt-injection checks (5) bill even without a document. Firecrawl pricing, billing docs.

What it means at scale, as arithmetic on 100,000 blocked pages:

100K blocked pages Spider Firecrawl
Target returns nothing usable At least 100,000 × 0.01 credit = 1,000 credits = $0.10, plus compute $0
Target returns a 403 page Bandwidth for each 403 page plus compute 100,000 credits, about a month of Standard ($99)

So Spider’s failure charge is tiny per page, and the real cost is the bandwidth of whatever the target sends back. Firecrawl’s is all-or-nothing: free when nothing comes back, a full credit when an error page does.

Firecrawl’s GitHub shows the edge cases. Issue #367 (closed September 2025) reported login-failed pages still being charged, and open issue #4212 (August 2026) says the Python SDK silently returns failed jobs, so check job status rather than trusting an empty result. Firecrawl issues.

Crawl defaults that can surprise your bill #

Spider’s crawl has no page cap by default; Firecrawl’s stops at 10,000 pages and refuses to start if you can’t pay for that many.

Spider’s API reference lists limit 0 as “no page cap”, depth 25, a 60-second request timeout (5-255) and respect_robots on. It also lists a crawl_timeout default of 2 minutes; I couldn’t tell from the reference how that interacts with a no-cap crawl, so I’d set limit explicitly rather than rely on it. Spider API.

On a per-byte bill, an uncapped crawl of a large site keeps metering until something stops it. Spider’s changelog does list a relevance gate (January 22, 2026) that skips irrelevant pages before billing when you crawl with an AI extraction prompt, which helps if you use one. Spider changelog.

Firecrawl’s /crawl defaults to a limit of 10,000 and returns a 402 if your credits can’t cover that limit. By default it follows only child paths of the starting URL, runs as an async job with signed webhooks, and keeps results for 24 hours. Crawl docs.

The 402 pre-check trips people up. Issue #2189 (closed) reported a 402 when no crawl limit was set, which fits the docs: on Free’s 1,000 credits, by my reading, you need a limit no higher than your balance. Firecrawl issues.

Both respect robots.txt by default. Spider lets you turn it off per request, while Firecrawl’s ignoreRobotsTxt is Enterprise-only. Spider FAQ, crawl docs.

Whichever you pick, pace the crawl; my web scraping rate limit guide covers how.

What are the rate limits on each? #

Spider gives every account the same high ceiling; Firecrawl scales limits with the plan. Verified 2026-11-17.

Limit Spider Firecrawl (Free / Hobby / Standard / Growth / Scale)
Requests a minute 10,000 core API requests per account, on every plan Scrape, map, search: 10 / 100 / 500 / 5,000 / 10,000
Crawl starts a minute Within the 10,000 Crawl and agent: 2 / 20 / 100 / 1,000 / 2,000
Concurrency concurrency_limit unlimited by default on pay-as-you-go 2 / 5 / 25 / 50 / 100 browsers
Queue Unlimited plan: “queues nothing server-side” 50K / 50K / 50K / 100K / 200K+ jobs, 48-hour maximum
Over the limit Unlimited plan: “immediate 429 with a Retry-After header” 429

Sources: Spider pricing, Spider FAQ, Spider API, Firecrawl rate limits.

How does Spider’s Unlimited plan work? #

Spider’s Unlimited plan is a different deal: you pay a flat monthly fee per concurrent slot instead of per byte. It runs from $40 a month for 2 slots ($20 each) through $350 for 25 ($14 each) to $19,994 for 1,538 ($13 each). Spider pricing.

Unlimited tier Per month Per slot
2 concurrent $40 $20
25 concurrent $350 $14
1,538 concurrent $19,994 $13

The fine print: Unlimited excludes LinkedIn and Facebook automation, Google search and multi-account use. Spider’s comparison post says Unlimited uses “standard proxy quality”, with higher-quality proxies on pay-as-you-go. Spider pricing, Spider’s comparison.

Is it worth it? $40 on pay-as-you-go buys 400,000 credits, which is 40 GB of bandwidth, or about 400K pages at an assumed 100 KB each. Unlimited only pays off if two always-busy slots move more than that in a month, and the pages I read give no per-slot throughput figure to check against.

The pricing page puts the self-serve ceiling at 1,538 slots, while the same page says “Need more than 1,500 concurrent? Talk to sales”. Neither matters below a few hundred slots.

What about JavaScript, proxies and output formats? #

Spider is cheaper per proxy feature and covers more countries; Firecrawl returns more formats per call.

Who renders JavaScript, and at what price? #

Spider’s default smart mode fetches over plain HTTP until a page needs JavaScript, then switches to Chrome; the pricing page also prices plain HTTP and full Chrome modes. Its pricing page lists JS rendering, proxy rotation and smart retry as included in every request, and its multipliers line reads “None for JS rendering or proxies”. Spider API, Spider pricing.

On Firecrawl, a page is 1 credit whether or not it needs rendering. Firecrawl also caches: maxAge defaults to 2 days, and a cached result still costs 1 credit. Scrape docs.

Which proxies and countries do you get? #

Spider’s proxy page lists residential proxies “From $1 / GB” in 199 countries, and the pricing page puts proxy mode at about $1-$4 per GB. Passing your own proxy with remote_proxy saves “50% on data transfer costs”. Spider proxies, Spider pricing, Spider API.

Here’s Spider’s third conflict. The API reference says stealth is on by default and routes to a residential pool when a target turns away datacenter IPs, and that residential pools “bill at 1.5–2× the flat-rate baseline”.

That doesn’t sit well with “None for JS rendering or proxies” on the pricing page. I can’t resolve it from public pages, so budget for the multiplier on protected targets.

Firecrawl has basic, enhanced and auto modes; auto, the default, retries with enhanced when basic fails, and enhanced adds 0 credits. Basic covers 26 listed countries and enhanced only the US and the Netherlands, and the proxies page doesn’t mention bringing your own. Firecrawl proxies, billing docs.

What does each one return? #

Spider returns raw HTML (the default), markdown, commonmark, text, XML or bytes, wrapped as JSON, JSONL, CSV or XML depending on the content type you ask for. Screenshots are a separate endpoint at about $0.0006 a request. Spider API, Spider pricing.

Firecrawl’s scrape call can return markdown, summary, html, rawHtml, screenshot, links, json, images, branding, product, audio, video and query answers, and it can run page actions before scraping. If you want structured fields, a screenshot and markdown from one call, that’s Firecrawl. Scrape docs.

For pure markdown quality, I compare converters in website to markdown for LLMs. Spider also added data connectors to S3, GCS, Google Sheets, Azure and Supabase in February 2026, which saves a step if your pipeline ends in one of those. Spider changelog.

Is Spider really that much faster? #

Spider says so, and I found no independent check. Every Spider speed number in circulation traces back to one test Spider designed and carried out itself.

Spider’s benchmark post (February 11, 2026, updated September 2) reports 74 pages a second for Spider against 16 for Firecrawl, about 4.6 times, and 99.9% success against 95.3%. On its 250 protected URLs, it reports 99.6% for Spider and 88.4% for Firecrawl. Spider’s benchmark.

What sits behind those numbers: 1,000 URLs (400 static, 350 single-page apps, 250 protected), “success” meaning an HTTP 200 with a non-empty body, a c6i.4xlarge instance in us-east-2, no stated concurrency and no linked dataset. Spider says plainly that it’s their product and they designed the test.

Spider’s comparison post (February 20, 2026, updated September 2) adds 182 against 27 pages a second on static pages and a retrieval score (recall@5) of 91.5 against 89.0. It also says the results are “not independently verified”. Spider’s comparison.

The independent data I found covers Firecrawl only:

Test Who ran it Firecrawl result Spider
Scrapeway, October 1 to November 1, 2026, 11 targets Scrapeway (nightly runs; “No affiliate links. No sponsors.”) 81% success, 5.2 s, $2.70 per 1,000 Not tested
Proxyway, December 2025, protected sites Proxyway, as reproduced by Zyte, a rival 33.69% at 2 requests/s, 26.69% at 10/s; “excellent crawler, poor unblocking API” Not tested

So on protected sites, Firecrawl has an outside score and Spider has only its own. That’s why I lean to Firecrawl for bot-protected targets, and why I’d trial both on your real URLs before committing either way.

Community signal is thin, and I couldn’t read Reddit. On Hacker News, one 2025 comment says Spider “has been really nice for crawling sites and saving the results as markdown”, and the Firecrawl complaint I found is price, in two comments (“Firecrawl is egregiously expensive”, September 2025). HN on Spider, HN on Firecrawl.

A pale concrete wall with a diagonal seam and round form-tie holes on the left, meeting a wall in near-total darkness on the right
Spider's speed numbers come only from tests Spider designed and ran itself. The outside tests I found, from Scrapeway and Proxyway, measured Firecrawl and left Spider out, so trial both on your own URLs before you rely on either. Photo: Rafael Garcin / Unsplash · Unsplash License

Is either one open source? #

Both have open-source code, under very different licences, and neither open version is the hosted product.

Spider Firecrawl
Repository spider-rs/spider, Rust firecrawl/firecrawl
Licence MIT “primarily licensed under” AGPL-3.0; SDKs and some UI MIT
GitHub stars About 2.9k About 202.4k
What’s open The crawler crate; the hosted backend isn’t shown Core scrape, crawl, map and search
What stays in the cloud The spider.cloud service Agent, Browser, interact, dashboards, enhanced proxies, enterprise controls

Spider’s crate is a Rust crawler that fetches over HTTP first and starts headless Chrome only when needed. Its README calls it “The fastest web crawler and scraper for Rust”, with no numbers attached. spider-rs.

Recent issues are crate bugs, such as #383 (with_delay() hanging a scrape, closed April 2026) and #384 (WARC output not replayable, closed April 2026), with no cloud-billing complaints in the latest 15. spider-rs issues.

In broad terms, MIT lets you embed and modify the crate in a closed product with little more than keeping the copyright and licence notice. That’s why it’s my pick for anyone building a crawler into something they sell, and one Hacker News commenter pointed to it “For those who want to self host.” HN on spider-rs.

Firecrawl’s server is AGPL-3.0. In broad terms, that’s a copyleft licence whose network clause asks you to share your changes if you modify the server and let others use it over a network. As I read it, calling the hosted API or using the MIT SDKs doesn’t raise that, but this is my general reading, not legal advice: if you plan to embed it, check with counsel.

Self-hosted Firecrawl is also a smaller product. The docs say it lacks the Fire-engine anti-bot layer, screenshots, actions, Agent, Browser, interact and some formats, and the stack it describes needs PostgreSQL, Redis, RabbitMQ and a Playwright service. Self-hosting docs, open source or cloud.

My Firecrawl self-hosting guide covers that setup, and open-source LLM crawlers compares the libraries.

What other comparisons get wrong #

Most pages ranking for this query carry at least one stale or wrong figure. These are the ones I checked against the vendors’ own pages.

Claims I checked on November 17, 2026

  • Spider compute at “$0.001/min” (TinyFish, mcp.directory): Spider’s pricing page says $0.0001 per CPU-minute, ten times lower.
  • Spider “failed requests are not billed” (TinyFish) or “cost nothing” (AIMultiple): Spider’s pricing FAQ now says “A failed page is not free”.
  • Firecrawl’s “Apache 2.0 core” (mcp.directory): the repository says AGPL-3.0, with MIT SDKs.
  • Spider at “$0 per month” (Respan): there’s no subscription, a one-time $0.25 of trial credits and a $5 minimum top-up. Respan also rates both 10.0/10 and lists Spider’s features as “Data not available”.
  • Firecrawl Hobby “3,000 credits” (Spider’s benchmark): the same post says 5,000 elsewhere, and Firecrawl’s pricing page says 5,000.
  • Scrapeway “August 2026: Firecrawl 64%, $5.18” (Spider’s comparison): I couldn’t see that month. The live Scrapeway table shows Firecrawl at 81% and $2.70 per 1,000, and $5.18 is ZenRows’ figure there.
  • Firecrawl at $16, $83 and $333 (mcp.directory): those are yearly-billing prices, unlabelled. Monthly is $19, $99 and $399.
  • Firecrawl “$3.80 per 1K JS-rendered requests” (AIMultiple): that’s Hobby’s $19 for 5,000 credits, true for that plan only.

The speed claims travel too. TinyFish repeats “~7x Firecrawl” as Spider’s claim, and mcp.directory cites “100k+ pages per second” as a best case; neither comes with an independent test.

Which one should you pick? #

Pick Spider if you’re crawling large numbers of light, mostly static pages, such as docs sites, blogs or a 100K-1M page RAG corpus. Spider’s own estimate for 1M pages is $28-$160 against $599-$749 on Firecrawl Scale, as long as your pages sit well under about 600 KB of metered bytes.

Pick Spider if your volume is small or irregular. A $5 top-up never expires, while Firecrawl’s cheapest paid route for 10K pages is $41-$44 every month. Below 1,000 pages a month, though, Firecrawl Free costs nothing.

Pick Firecrawl if you need structured JSON, screenshots, page actions or search from one API. Standard at $99 covers 100K pages or 20K JSON pages, with nine official SDKs.

Pick Firecrawl if your pages are heavy with JavaScript or media, or finance wants a fixed number. It’s 1 credit per page regardless of weight, with a cap on extra spend, while Spider’s Chrome metering is unpublished.

Pick Firecrawl for bot-protected targets, on the only independent evidence there is: 81% success on Scrapeway in October 2026. No third party has tested Spider, so trial both on your own URLs before a big commitment.

Pick Spider’s spider-rs crate if you’re embedding or modifying a crawler in a product you sell. MIT is far simpler to live with than Firecrawl’s AGPL server.

What I’d do first #

Pull 100 URLs from your real target list and look up their average raw HTML size; that’s the number Spider meters. Under about 0.6 MB a page, I’d start on Spider’s $5 top-up; above it, or if you need JSON extraction, I’d start on Firecrawl Free and move to Standard.

Either way, set an explicit crawl limit on day one. My guide to reducing web scraping costs has more ways to trim the bill.

What else is worth a look? #

FAQ #

Is Spider cheaper than Firecrawl? #

For light pages, by a wide margin on list prices. Spider’s own estimate for 1M pages is $28-$160, against $749 a month ($599 billed yearly) on Firecrawl Scale. Spider bills $1 per GB, though, so once your pages average more than roughly 0.6-1 MB of metered bytes, Firecrawl’s flat 1 credit a page comes out cheaper. Spider pricing, Firecrawl pricing.

Does Spider charge for failed pages? #

Yes, according to its current pricing FAQ: “A failed page is not free.” Each attempt bills bandwidth and compute whatever status comes back, with a 0.01-credit minimum for a blocked or empty page. Older third-party pages, and one line in Spider’s own guide, still say failures are free. Spider FAQ.

Do Firecrawl credits roll over? #

Not on Free, Hobby, Standard or Growth: plan credits reset every month, even on yearly billing. Firecrawl’s docs say annual Scale rolls over one month and annual Enterprise two. Extra credits you buy expire if you cancel your subscription. Billing docs.

Is Spider open source? #

The crawler is. spider-rs is an MIT-licensed Rust crate with about 2.9k GitHub stars. The hosted spider.cloud API is a paid service, and its backend code isn’t in that repository as far as I could see. spider-rs.

Is Firecrawl open source, and what does AGPL mean for me? #

Its core is, under AGPL-3.0, and its SDKs are MIT. In broad terms, calling the hosted API or using the SDKs doesn’t touch the AGPL; modifying the server and offering it to others over a network is where its share-your-changes terms apply, so check with counsel before embedding it in a product. Self-hosted Firecrawl also lacks the cloud’s Fire-engine anti-bot layer, screenshots and actions. Firecrawl repository, self-hosting docs.

How fast is Spider compared with Firecrawl? #

Spider’s own February 2026 benchmark reports 74 pages a second against Firecrawl’s 16, about 4.6 times, on 1,000 URLs. Spider designed and carried out that test, didn’t state the concurrency and links no raw data. I found no independent test that includes Spider. Spider’s benchmark.

Does Spider have a free tier? #

Not a monthly one. Spider’s FAQ says new accounts get 2,500 free credits, worth $0.25, if the signup passes its checks, and there’s no subscription; after that you top up from $5. Firecrawl Free gives 1,000 credits every month with no card. Spider FAQ, Firecrawl pricing.

Sources

30 sources, accessed .

Official pricing, docs, changelogs and status pages 17

  1. Spider pricing spider.cloud · accessed
  2. Spider FAQ spider.cloud · accessed
  3. Spider guide: pricing and plans spider.cloud · accessed
  4. Spider API reference spider.cloud · accessed
  5. Spider residential and ISP proxies spider.cloud · accessed
  6. Spider changelog spider.cloud · accessed
  7. Spider vs Firecrawl (Spider's own comparison) spider.cloud · accessed
  8. Firecrawl vs Crawl4AI vs Spider benchmark (run by Spider) spider.cloud · accessed
  9. Firecrawl pricing (markdown version) firecrawl.dev · accessed
  10. Firecrawl billing docs docs.firecrawl.dev · accessed
  11. Firecrawl rate limits docs.firecrawl.dev · accessed
  12. Firecrawl crawl docs docs.firecrawl.dev · accessed
  13. Firecrawl scrape docs docs.firecrawl.dev · accessed
  14. Firecrawl proxies docs docs.firecrawl.dev · accessed
  15. Firecrawl self-hosting docs docs.firecrawl.dev · accessed
  16. Firecrawl: open source or cloud docs.firecrawl.dev · accessed
  17. Firecrawl changelog firecrawl.dev · accessed

GitHub code, issues and releases 4

  1. spider-rs repository github.com · accessed
  2. spider-rs issues github.com · accessed
  3. Firecrawl repository github.com · accessed
  4. Firecrawl issues about crawl credits github.com · accessed

Independent benchmarks and published studies 1

  1. Scrapeway monthly scraping API benchmark scrapeway.com · accessed

Review platforms 2

  1. Respan market map: Firecrawl vs Spider respan.ai · accessed
  2. AIMultiple: Firecrawl alternatives aimultiple.com · accessed

Community discussions 3

  1. Hacker News comments mentioning spider.cloud hn.algolia.com · accessed
  2. Hacker News comments on Firecrawl's price hn.algolia.com · accessed
  3. Hacker News mentions of spider-rs hn.algolia.com · accessed

Other blogs 3

  1. Zyte's summary of Proxyway's December 2025 test zyte.com · accessed
  2. TinyFish: Firecrawl alternatives tinyfish.ai · accessed
  3. mcp.directory five-crawler comparison mcp.directory · accessed

What readers say

Used one of the tools in this guide? Tell other readers how it went. I read every review before it goes up, and I never turn one down for being negative. How reviews work.

No reader reviews yet. Yours could be the first.

Write a review
Your rating

What did you use it for, what worked, and what didn't? 30 to 1500 characters. Please don't include links.

Your connection to the company

Your review appears with the name you enter, after I approve it. I don't ask for your email. Privacy.

  1. Explainer

    How to crawl an entire website for your LLM

    Find every URL, keep the crawl in scope and price it first. Settings and costs for Firecrawl, Apify, Crawl4AI, Cloudflare and Spider, checked Nov 2026.

    14 min read