Firecrawl Pricing 2026: API Cost, Limits, Credits Usage

Firecrawl's headline price is one credit per page, which makes it look simple. It isn't. Surcharges stack on top of every request, top-up credits cost up to 2.5 times the plan rate, and an X post costs thirty credits. This is the full breakdown: every tier, every endpoint, every surcharge, and the exact points where upgrading beats topping up.

Firecrawl pricing at a glance

Firecrawl has a free tier plus four paid plans and custom Enterprise. Free gives 1,000 credits, Hobby is $19 for 5,000, Standard $99 for 100,000, Growth $399 for 500,000, and Scale $749 for 1,000,000. Annual billing takes two months off every tier. The free plan runs at 2 concurrent browsers and 10 requests a minute — enough to evaluate, not to produce.

The shape here is a cliff, not a slope. Hobby to Standard is 20 times the credits for 5.2 times the price, dropping the unit rate from $3.80 to $0.99 per thousand. After that the curve flattens hard — Growth improves on Standard by only 19%, and Scale improves on Growth by 6%. Almost all the value in upgrading is captured in that first step.

What each endpoint costs

A credit is not a page. It is a unit of work, and different endpoints consume it differently. Scrape and crawl cost 1 credit per page, map costs 1 credit per call, search costs 2 credits per 10 results, and interact costs 2 to 7 credits per browser minute.

/map being billed per call rather than per page is the most exploitable fact in this table. Discovering every URL on a large site costs a single credit, so mapping first and scraping selectively is dramatically cheaper than crawling everything and filtering afterwards.

The surcharges that change everything

Base rates are only the starting point. Options stack on top of every page, and this is where most Firecrawl bills are actually made:

The advertised rate is for the least useful output format. Budget from the one you will actually ship.

Two consequences follow. First, the free tier's 1,000 credits is only 200 pages if you use JSON mode — worth knowing before you conclude the free tier is generous. Second, a Standard plan's 100,000 credits becomes 20,000 extracted pages, which moves your effective rate from $0.99 to $4.95 per thousand.

Rate limits and concurrency

Credits cap how much you can do in a month; rate limits cap how fast. They scale independently of each other, and on Firecrawl they scale steeply — Growth gets ten times Standard's scrape rate, which is a much bigger jump than the credit allowance alone suggests.

Note that /interact is the one limit that does not keep pace at the top — Scale gets 1,500 per minute against Growth's 1,000, a far smaller step than the other endpoints. If browser interaction is your bottleneck, Scale buys you less than the price difference implies.

Pay-as-you-go: top-ups cost more

All paid plans support auto-recharge, which adds credits when your balance runs out, in $5 increments. You can cap monthly spend or disable it entirely. What is not obvious is that top-up credits are priced well above your plan's own rate.

Standard and Growth users pay two and a half times their plan rate for overflow. That is the single most avoidable cost on the platform, and it punishes exactly the mid-tier users most likely to run a variable workload.

When to upgrade instead of topping up

The arithmetic gives you two clean thresholds:

• Standard to Growth: 220,000 credits a month. Below that, Standard plus auto-recharge is cheaper. At 220,000 they cost the same $399. Above it, Growth wins and keeps winning.
• Growth to Scale: 675,000 credits a month. Below that, Growth plus top-ups beats Scale's $749. Above it, Scale is cheaper.

Two caveats make the real thresholds lower than the maths suggests. Rate limits jump enormously between tiers, so if you are throttled rather than out of credits, upgrade regardless. And credits do not roll over below Scale, so a spiky workload that under-uses some months and overflows others gets penalised at both ends.

Rollover, expiry and cancellation

Only Scale carries unused credits forward, for one month; Enterprise terms are custom. Hobby, Standard and Growth credits expire at the end of the billing cycle. Separately, purchased top-up credits expire when you cancel your subscription — so buying a large block on the way out of a project loses you whatever is left.

The practical implication is to size your plan to a typical month rather than your worst one, and absorb the peaks with capped auto-recharge. Paying for a bigger tier to cover an occasional spike means burning unused credits every other month.

Monthly versus annual

Annual billing is a flat two months free — about 16.7% — on every tier: $38 a year on Hobby, $198 on Standard, $798 on Growth and $1,798 on Scale. There is no tier-shopping advantage, since the percentage never changes.

Because the discount is uniform, the only question is commitment. Run a month on monthly billing, measure your real credit burn including surcharges, then commit annually to whichever tier that number implies. On Standard that costs you about $16 to learn the answer.

Six ways to cut your Firecrawl bill

1. Map before you crawl. /map costs 1 credit per call and returns every URL. Scraping a filtered list beats crawling a whole site and discarding most of it.
2. Drop JSON mode where you can. Markdown at 1 credit plus your own extraction pass often beats 5 credits a page, particularly if a model is already in your pipeline.
3. Only enable ZDR where compliance requires it. It is a permanent +1 credit on every page, which is a 100% surcharge on plain scrapes.
4. Watch the interact prompt. Using a prompt takes interaction from 2 to 7 credits per browser minute — more than triple.
5. Cap auto-recharge. At 2.5x the plan rate on Standard and Growth, uncapped overflow is the most expensive way to buy credits.
6. Avoid scraping X at volume. At 30 credits a post, a thousand posts costs $29.70 on Standard — the same as thirty thousand ordinary pages.

Worked examples: three real projects

Abstract rates are hard to reason about, so here are three concrete workloads costed end to end on the Standard plan at $0.99 per thousand credits.

1. Ingesting a documentation site for RAG — 4,000 pages, markdown only. One /map call to discover URLs (1 credit), then 4,000 scrapes at 1 credit each. Total 4,001 credits, about $3.96. This is the workload Firecrawl is cheapest at, and it fits inside the free tier's monthly allowance four times over if you split it.

2. Extracting structured product data — 10,000 pages with JSON mode. Ten thousand pages at 5 credits each is 50,000 credits, about $49.50 — half a Standard plan for a single job. Scraping to markdown first (10,000 credits, $9.90) and running your own extraction pass costs $9.90 plus your model tokens, which is usually cheaper unless the schema is genuinely complex.

3. Daily monitoring of 500 competitor pages. Monitoring bills 1 credit per page when checks error, so a clean run costs little and a bad day costs up to 500 credits. Budget the worst case at 500 credits a day, roughly 15,000 a month, about $14.85 — comfortably inside Standard, and a good argument for not over-buying a tier.

The pattern across all three: plain scraping is almost free at these volumes, and structured extraction is the only thing that meaningfully moves the bill. If a project looks expensive, the JSON surcharge is nearly always the reason.

Self-hosting as a pricing lever

Firecrawl is open source and self-hostable via Docker, and the open build covers the core scraping and crawling engine plus Map, Search, batch scrape, Extract, JSON mode, change tracking and the SDKs. Running it yourself means no credits at all — you pay for servers and bandwidth instead.

That changes the calculation at high volume. A workload spending $749 a month on Scale is buying a million credits; the same work self-hosted costs whatever a few browser containers cost you, which is typically far less at that scale. What you give up is managed infrastructure, the higher concurrency of the hosted tiers, and LLM extraction applied during the crawl.

The honest guidance: self-host if you are already comfortable operating containers and your volume has pushed you toward Growth or Scale. Stay hosted below that, because at Standard's $99 a month the ops time costs more than the credits do.

Which plan should you pick?

• Evaluating — Free, 1,000 credits. Remember that is only 200 pages with JSON extraction.
• Small, occasional jobs — Hobby at $19. Accept that you are paying $3.80 per thousand pages, nearly four times Standard's rate.
• Any production workload — Standard at $99. The unit rate collapses here and this is where most teams should start.
• Above roughly 220,000 credits a month, or throttled at 500 requests a minute — Growth at $399.
• Above roughly 675,000 credits a month, or you need credit rollover — Scale at $749.
• SSO, zero data retention across the org, or custom API restrictions — Enterprise.

The verdict on Firecrawl's pricing

Firecrawl's pricing is fair but front-loaded with traps for the unwary. The base rate is genuinely competitive, the annual discount is honest, and /map being per-call is quietly generous. Against that: the format most people need costs five times the advertised rate, overflow costs two and a half times the plan rate, and credits vanish monthly below Scale.

None of that is hidden — it is all in the billing documentation — but none of it is on the pricing page's headline either. Model your real workload, in the real output format, before choosing a tier. For what the platform actually does with those credits, see our full Firecrawl review, or compare it against a compute-metered alternative in Apify vs Firecrawl.

Start on Firecrawl free →

1,000 free credits monthly, no card · Paid plans from $19/month · Cancel anytime