Firecrawl is the scraping API most AI teams try first. It turns a URL into clean markdown that a language model can read, it can crawl a whole site from one starting link, and it has a search endpoint that returns results and page content in one call. The pricing page lists credits per month, and a credit sounds like a page. For the most common AI use, it is not.
The plans and credit rates below were read from firecrawl.dev/pricing and the billing page of the Firecrawl docs on 30 September 2026. The pricing page defaults to yearly billing, so the first numbers you see are the discounted ones. Check the live page before you sign, because this market reprices several times a year.
The Firecrawl plans
Five self serve plans, including a free one, plus Enterprise through sales. Credits reset every month and do not roll over, except on yearly Scale plans, which carry unused credits forward for one month. When credits run out, paid plans can buy more in 5 USD blocks up to a monthly limit you set, at a rate that depends on the plan.
| Plan | Monthly | Yearly, per month | Credits | Concurrent | Extra credits per 5 USD |
|---|---|---|---|---|---|
| Free | 0 | 0 | 1,000 | 2 | none, requests stop |
| Hobby | 19 USD | 16 USD | 5,000 | 5 | 1,000 |
| Standard | 99 USD | 83 USD | 100,000 | 25 | 2,000 |
| Growth | 399 USD | 333 USD | 500,000 | 50 | 2,500 |
| Scale | 749 USD | 599 USD | 1,000,000 | 100 | 5,000 |
The overage column deserves a look before you pick a plan. Extra credits always cost more than plan credits. On Hobby they are 5 USD per 1,000 against 3.80 inside the plan. On Standard and Growth the gap is widest, 2.50 and 2.00 USD per 1,000 against 0.99 and 0.80, about two and a half times the plan rate. So a team that tops up every month should move up a plan rather than keep buying blocks.
What one Firecrawl credit buys
The base rate is simple. Scrape, crawl and map cost 1 credit per page, and the enhanced proxy mode that Firecrawl switches to for harder sites costs nothing extra. Search costs 2 credits per 10 results, rounded up, plus the scrape cost of any result you ask it to fetch.
The options are where the bill moves. From the docs billing page:
| Option | Extra credits per page | Total per page |
|---|---|---|
| Markdown or HTML only | 0 | 1 |
| JSON format, LLM extraction | 4 | 5 |
| JSON plus zero data retention | 5 | 6 |
| JSON plus prompt injection check | 8 | 9 |
| JSON plus PII redaction | 8 | 9 |
| PDF parsing | 1 per PDF page | 2 per PDF page |
The modifiers stack, and they apply inside crawl and search too, because both scrape each page they return. One more detail from the same page: a target that answers 403 Forbidden or 404 Not Found still counts as a returned document and costs 1 credit. Firecrawl only waives the charge when it returns nothing at all.
Firecrawl cost per 1,000 pages by plan
Plan fee divided by credits, times credits per page, on monthly billing:
| Plan | Markdown page | JSON page | 1,000 searches, results only |
|---|---|---|---|
| Hobby, 19 USD | 3.80 USD | 19.00 USD | 7.60 USD |
| Standard, 99 USD | 0.99 USD | 4.95 USD | 1.98 USD |
| Growth, 399 USD | 0.80 USD | 3.99 USD | 1.60 USD |
| Scale, 749 USD | 0.75 USD | 3.75 USD | 1.50 USD |
For comparison, our plans charge one successful request per page whatever you ask for: 0.98 USD per 1,000 on Starter, 0.60 on Growth and 0.33 on Scale on monthly billing. Billed yearly, which is half price on every plan, that becomes 0.48, 0.30 and 0.17. A Google results page is one request at the same rate.
The JSON multiplier, or why 500,000 pages can mean 100,000
Growth advertises 500,000 pages scraped for 399 USD. That is true for markdown. Most AI teams we talk to do not want markdown, though. They want fields: a price, a job title, a company name, a date, typed and ready for a database. On Firecrawl that is the JSON format, and at 5 credits per page the same 399 USD buys 100,000 structured pages.
None of the review articles ranking for Firecrawl pricing make this conversion, and it is the number that decides the budget. If half your pages need JSON, Growth covers about 167,000 pages a month, not 500,000. Add zero data retention, which many companies require before they send customer related URLs to a vendor, and a JSON page is 6 credits.
On our side, named field extraction is part of the request. Starter takes a CSS or XPath schema and returns JSON. Growth adds plain English extraction, where you describe the fields and get typed JSON back, and it costs the same 0.60 per 1,000 as a page with no extraction. Both approaches work. They are just priced very differently once extraction is the default rather than the exception, which the AI web scraper page covers in more detail.
Real budgets, Firecrawl against flat per request pricing
Six monthly workloads, each priced at the cheapest Firecrawl plan (plus top ups where that is cheaper) and the cheapest of our plans that covers it. Monthly billing on both sides.
| Monthly workload | Credits needed | Firecrawl | Flat per request |
|---|---|---|---|
| 50,000 markdown pages | 50,000 | 99 USD, Standard | 49 USD, Starter |
| 100,000 markdown pages | 100,000 | 99 USD, Standard | 149 USD, Growth |
| 50,000 pages as JSON | 250,000 | 399 USD, Growth | 149 USD, Growth |
| 200,000 pages as JSON | 1,000,000 | 749 USD, Scale | 149 USD, Growth |
| 1,000 searches a day, top 5 results as JSON | 810,000 | 749 USD, Scale | 149 USD, Growth (180,000 requests) |
| 1,000,000 markdown pages | 1,000,000 | 749 USD, Scale | 499 USD, Scale |
The second row goes to Firecrawl, and we would not argue with it. For 100,000 plain pages a month, Standard at 99 USD beats our 149. The JSON rows go the other way by three to five times, because the extraction multiplier lands on every page. The search row is the typical research agent pattern: 2 credits for the results and 5 for each page it reads, 27 credits per question.
Where Firecrawl is the right buy
Markdown for a language model. If you feed whole pages into a prompt or a vector store, Firecrawl's markdown output is its core product and it is good at it. At 0.75 to 0.99 USD per 1,000 on Standard and up, it is priced in line with the market for that job.
Crawling and mapping whole sites. The crawl and map endpoints discover and fetch every page from one starting link. If you do not know the URLs in advance, that saves you writing a crawler. Our website crawler API covers the crawl part with JSON per page, but has no map endpoint and caps one crawl at 2,000 pages on Scale.
Small budgets and concurrency. Hobby at 19 USD is a cheaper entry point than our 49 USD Starter, and Standard's 25 concurrent requests are five times our Starter's 5. The open source version can also run on your own servers if a vendor is not an option at all.
Where flat per request pricing wins
Structured data at volume. When every page has to come back as fields, a flat price per page is about three to five times cheaper at the same volume, as the table shows. That is most price monitoring, lead enrichment, job and listing feeds.
Protected targets. Firecrawl charges 1 credit for a page that comes back as a 403, so a crawl that hits a wall of refusals still spends credits. Our plans bill only successful responses, and pages behind Cloudflare or DataDome cost the same as any other.
Agents that search and read. Pairing search with page reads is where credit math compounds fastest. A Google results page is one request on our SERP API, and each page the agent reads is one more. If those pages end up in a model's context, it is also worth running them through a prompt injection filter before the agent acts on them, whichever scraper fetched them.
Most teams do some of both. Run a day of your real targets through Firecrawl's free credits and our plans, count how many pages needed JSON, and price the month both ways.
Questions buyers ask
How much does Firecrawl cost per month? Hobby is 19 USD, Standard 99 USD, Growth 399 USD and Scale 749 USD on monthly billing, for 5,000, 100,000, 500,000 and 1,000,000 credits. Billed yearly they drop to 16, 83, 333 and 599 USD a month. Enterprise is priced by sales, and a free plan gives 1,000 credits a month.
How many credits does a Firecrawl scrape cost? One credit per page for markdown or HTML, including crawl and map. Structured JSON extraction adds 4 credits, zero data retention 1, and PII redaction or a prompt injection check 4 each. The modifiers stack, so a JSON page with PII redaction costs 9 credits.
Is Firecrawl free? There is a free plan with 1,000 credits a month, 2 concurrent requests and no card. It is enough to test your targets. It has no pay as you go, so requests return an error once the month's credits are used.
Does Firecrawl charge for failed requests? Only when it returns no document. A page that answers 403 or 404 is returned to you and costs 1 credit, so blocked pages are billed. Our plans bill only successful responses, so a refusal costs nothing, which is worth knowing if your targets push back.
Do unused Firecrawl credits roll over? Not on Hobby, Standard or Growth, and not on monthly Scale. Yearly Scale plans carry unused plan credits over for one month. Plan the plan size around a normal month, then top up for spikes.
Is Firecrawl cheaper than ScrapingBee or ZenRows? For markdown pages, often yes, because rendering and enhanced proxies do not add credits the way they do on ScrapingBee or ZenRows. For JSON extraction the 5 credit page closes most of that gap. Price your own mix of pages, not the headline. Browserbase, which bills browser hours and proxy gigabytes instead of credits, gets the same treatment in our Browserbase pricing article.