9 min read
Firecrawl costs $19 a month on Hobby, $99 on Standard, $399 on Growth and $749 on Scale, with a free tier of 1,000 credits. Every plan buys credits, and a plain scrape is 1 credit a page. The number that decides your real bill is the one most summaries skip: asking for structured JSON adds 4 credits per page, so a page you actually want as clean fields costs 5 credits. Standard is 100,000 plain pages or about 20,000 structured ones.
figures taken from firecrawl.dev/pricing and its published plan matrix
Robot console
Standing by Running · s Complete · rows
1 · Pick a target
3 · Fields to extract
Agent log
Crawl graph
Extracted data · rows
Want this data fresh every morning, without lifting a finger?
The numbers
| Plan | Billed monthly | Billed annually | Paid up front for a year | Credits a month | Concurrent browsers | Support |
|---|---|---|---|---|---|---|
| Free | $0, no card required | 1,000 | 2 | Community | ||
| Hobby | $19/mo | $16/mo | $190 (saves $38) | 5,000 | 5 | Basic |
| Standard | $99/mo | $83/mo | $990 (saves $198) | 100,000 | 25 | Standard |
| Growth | $399/mo | $333/mo | $3,990 (saves $798) | 500,000 | 50 | Priority, shared Slack |
| Scale | $749/mo | $599/mo | $7,190 (saves $1,798) | 1,000,000 | 100 | Priority, invoiced billing |
| Enterprise | Custom credits and terms, SSO and SCIM, zero data retention, a support SLA and a custom analytics dashboard | |||||
The annual discount is 16.7 percent on Hobby, Standard and Growth, and 20 percent on Scale. Credit rollover and invoiced billing only start at Scale, so on Hobby through Growth an unused month is money gone. If your workload is seasonal, that argues for staying monthly and topping up rather than prepaying a year.
The meter
| Endpoint or option | Credit cost | What that means in practice |
|---|---|---|
| Scrape | 1 credit | Per page, returned as markdown or HTML |
| Crawl | 1 credit | Per page discovered and fetched during the crawl |
| Map | 1 credit | Per call, to list the URLs on a site |
| Search | 2 credits | Per 10 results; scraping those results bills separately |
| Monitor | 1 credit | Per page, per check, so an hourly watch on one page is 720 credits a month |
| Interact | 2 credits | Per browser minute in a sandbox session |
| JSON format | +4 credits | Per page, on top of the scrape or crawl credit. This is the structured extraction fee |
| Prompt injection check | +4 credits | Per page, on top of the scrape credit |
| PDF parsing | 1 credit | Per PDF page, which makes long documents expensive |
| Agent (preview) | 5 free runs a day | Then dynamic pricing based on agent tokens |
The 5x rule
Nobody buys a scraping API to store markdown. You want price, title, SKU and stock as fields, and that means the JSON format, which is 1 credit plus 4. Every plan allowance in the marketing table should be divided by five before you compare it with anything else. Standard is 20,000 structured pages a month, Growth is 100,000, Scale is 200,000.
The blocked-page rule
Firecrawl does not charge when a scrape returns nothing at all, but a page that answers with an error status such as 403 or 404 is handed back to you and billed as a delivered response. On a target list with heavy bot defence, a meaningful slice of your credits can go to pages you never used. Budget for a failure rate rather than assuming every credit becomes a row.
Overage
| Plan | Credits per $5 increment | Effective rate per 1,000 credits | Credits roll over |
|---|---|---|---|
| Free | Not available | Not available | No |
| Hobby | 1,000 | $5.00 | No |
| Standard | 2,000 | $2.50 | No |
| Growth | 2,500 | $2.00 | No |
| Scale | 5,000 | $1.00 | One month |
Compare those overage rates with the plan rates and the upgrade path is obvious. Standard gives 100,000 credits for $99, which is $0.99 per 1,000, while Standard overage is $2.50 per 1,000. Overage is two and a half times the in-plan price, so a month that regularly spills over is a month you should have spent on the next tier. You can cap the damage: set a whole-dollar monthly pay-as-you-go limit in billing settings, enter 0 to switch it off entirely, or leave it blank for no ceiling at all. Leaving it blank on a job with a runaway crawl is how surprise invoices happen.
Real months
Worked bill 01
A documentation corpus refreshed monthly, taken as markdown rather than fields, is 3,000 credits. Hobby at $19 covers it with room to spare, and the free tier covers a third of it. For this shape of work Firecrawl is very cheap, and that is exactly what it was built for.
Worked bill 02
Four runs a month is 40,000 pages, and with JSON extraction that is 200,000 credits. Standard's 100,000 is short, so this lands on Growth at $399 a month, or $333 billed annually. Add the developer who writes the schema, the scheduler and the retry logic, because none of that is in the price.
Worked bill 03
Monitor bills 1 credit per page per check, so 300 pages checked every hour is 216,000 credits a month before you extract anything from the pages that changed. That is Growth territory at $399, and the diffing, alert routing and history are still yours to build.
The other half
Not included 01
You describe the fields you want in a JSON schema, in code, and you maintain it when the source changes. A robot you set up by typing a sentence is a different purchase from an endpoint you write against.
Not included 02
Firecrawl runs when you call it. Cron, back-off, deduplication against last week's rows and a record of what changed are your infrastructure. That work is invisible on the invoice and very visible on the engineering roadmap.
Not included 03
Supplier portals, member directories and dashboards need a sign-in, a search and pagination on every run. The Interact endpoint gives you a browser sandbox at 2 credits a minute, but the flow itself is something you script and keep working.
Not included 04
The response is JSON to your code. Getting it into a Google Sheet a merchandiser reads, or a Slack channel that pings when a price moves, is another service you build and host.
One more line item is easy to miss when you pipe scraped pages into a model. Web content is untrusted input, and a page can carry instructions aimed at whatever agent reads it next, which is why Firecrawl sells a prompt injection check at 4 credits a page. If that content reaches a tool-using agent, the guardrails that sit between an agent and its tools belong in the budget alongside the scraping credits.
Comparison
| Tool | What the meter counts | Entry paid price | You still build |
|---|---|---|---|
| Firecrawl | Credits: 1 per page, 5 per structured page | Hobby $19/mo, Standard $99/mo | Schema, scheduling, retries, delivery |
| Apify | Compute units, proxy traffic and per-result Actor fees | Starter $19/mo plus usage | The Actor, or the trust that a rented one keeps working |
| Zyte | Successful requests, priced by site difficulty tier | Pay as you go from $0.13 per 1,000 HTTP requests | Spiders, parsing, orchestration |
| Browse AI | Credits: 1 credit per 10 rows, 1 per detail page | Personal $48/mo, $19 billed annually | Little, within its website and robot caps |
| WebRobot | Records delivered, counted once whether one page or two filled them | Launch $79/mo, $63 billed yearly, 10,000 records | Nothing: the robot, the schedule, the delivery and the API are the product |
Read that table by asking who is doing the work. If you have engineers and want an endpoint, Firecrawl is priced well and the free tier is a real one. If you want rows on a schedule and nobody on the team wants to own a scraper, the comparison that matters is the Firecrawl alternative comparison, and the same trade-off shows up in Apify pricing and Browse AI pricing. For the whole market at once, the cost of web scraping breaks down build against buy.
FAQ
Describe the data in plain English. The robot runs on a schedule and delivers 10,000 records a month to Sheets, Slack or your API for $79.