Firecrawl Pricing Explained (2026): Credits, Tiers, and the Hidden Extract Bill

By the fastCRW team · Pricing verified 2026-05-18 · Vendor pricing is volatile — confirm on firecrawl.dev/pricing before deciding.

Disclosure: This is written by the fastCRW team. fastCRW is a Firecrawl-compatible alternative, so we are not neutral. Every Firecrawl number below is sourced from their public pricing page or first-party docs as of 2026-05-18; we have tried to be accurate, and we tell you where numbers are in dispute.

The short version

Firecrawl is a well-funded, polished web-data API for AI agents. Its pricing looks simple on the surface — a per-page credit model with a few subscription tiers — but two details catch teams off guard: the free tier is a one-time grant, not a monthly allowance, and structured extraction can run on a separate token-based subscription stacked on top of your plan. This post walks every tier, every credit rate, and the two traps, then shows the math at realistic volumes.

The tier table

As of 2026-05-18, Firecrawl's published tiers (annual-billed effective pricing) look like this:

One nuance worth flagging: Firecrawl's own pricing page shows Standard at $83/mo on annual billing, but several third-party price trackers cite $99/mo as the monthly-list rate. Both numbers are in circulation. Treat $83 as the annual-effective figure and $99 as month-to-month list. We use $83 below to be charitable.

How the credit model works

Firecrawl meters in credits, and the headline mapping is friendly:

• Scrape: 1 credit per page.
• Crawl: 1 credit per page crawled.
• Search: roughly 2 credits per 10 results.
• Interact / browser actions: ~2 credits per browser-minute.

For plain scrape and crawl, "1 credit = 1 page" is honest and easy to forecast. The complications start when you add structured extraction or render-heavy pages — modes that can consume more than one credit per page depending on configuration. Always read the response metadata to see how many credits a call actually billed; don't assume the headline rate.

Trap 1: the free tier is one-time, not monthly

This is the single most common surprise. Firecrawl's free tier grants 1,000 credits once — a lifetime allowance, not a monthly refill. A developer who runs a 500-page crawl on Monday and a 400-page crawl on Tuesday has burned 90% of their free quota and will hit a wall by Wednesday with no monthly reset to wait for. Many trial users exhaust the entire grant in a single evaluation session and conclude the product "ran out" before they finished testing.

This matters when you compare free tiers across vendors. A "1,000 one-time" grant and a "1,000 per month" grant are wildly different economically even though the number on the page is identical. Read the word "one-time" carefully on any pricing page — it changes everything.

Trap 2: extract can be a second subscription

This is the bigger one for AI teams. Firecrawl's AI-powered structured extraction (turning a page into JSON against a schema) is widely reported to run on a separate token-based subscription rather than drawing from your main credit pool. The practical consequence: a team on the Standard plan ($83–99/mo) that also needs extraction reportedly pays Standard plus an extraction add-on of roughly $89/mo — a combined ~$172–188/mo minimum before any usage multipliers.

If your use case is RAG ingestion of clean markdown, you may never touch extract and this never bites you. But "turn this page into structured JSON" is exactly what most agent and enrichment pipelines need, so the dual-billing structure tends to surface precisely for the workloads that adopt the product hardest. Several public HN and Reddit threads describe the combined bill as "egregiously expensive." Budget for it before you commit.

What it actually costs at scale

Let's price three realistic workloads on Standard ($83/mo, 100,000 credits/mo):

• Small RAG bot: 2,000 pages/day = 60,000 credits/mo. Fits inside Standard. Effective cost: $83/mo. Fine.
• Growing agent product: 5,000 pages/day = 150,000 credits/mo. Over Standard's 100k; you move to Growth at $333/mo or buy overage. Cost jumps ~4x for a 2.5x volume increase.
• Enrichment pipeline with extraction: 3,000 pages/day with JSON extraction on each. Now you're paying Standard + the extract subscription, so the floor is ~$172–188/mo before you even check whether your volume fits the credit cap.

The pattern: plain scraping is predictable and reasonably priced; the cost surprises live in the free-tier expiry, the extract dual-billing, and the step-function jumps between tiers when you cross a credit ceiling.

The escape hatch most pricing pages don't have

Here is the structural point. Every cost above assumes you must use a hosted, metered service. fastCRW is built differently: it is an open-core Rust scraper. The engine is a single ~6MB binary licensed under AGPL-3.0 that you can self-host with unlimited requests and zero license fee, exposing a Firecrawl-compatible API (/v1/scrape, /v1/crawl, /v1/map, /v1/search). There is also a Managed Cloud if you'd rather not run infrastructure.

What this changes about the pricing conversation: your worst-case cost is capped at the price of a small VPS, because you can always run the same engine yourself. You are never trapped by a tier jump or an add-on subscription, because the floor underneath you is software you already control.

fastCRW's own tiers: a recurring free tier with a one-time lifetime grant of 500 credits (and a free unlimited local self-host mode), Hobby at $13, Standard at $69 for 100k credits, Growth at $279 for 500k, Scale at $549 for 1M. Tier-for-tier these undercut Firecrawl, and crucially there is no separate extract subscription — structured JSON extraction is part of the same scrape call under the same credit model.

How to forecast your real Firecrawl bill

1. Count pages, not requests. A crawl of a 5,000-page site is 5,000 credits, not 1. Map the site first to get an accurate page count.
2. Decide if you need extract. If yes, assume the dual-subscription floor (~$172–188/mo on Standard). If no, the headline tier price is closer to reality.
3. Find your tier ceiling. Divide monthly pages by the tier credit cap. If you're within ~70% of a cap, plan for the next tier — usage is rarely flat.
4. Model the worst month, not the average. Spiky agent traffic blows through credit caps fast; the overage or forced tier upgrade is where surprise bills come from.
5. Price the self-host floor. Whatever the hosted number is, compare it to "the same workload on a $5–12/mo VPS with an open-core engine." That delta is your negotiating leverage and your risk cap.

So is Firecrawl's pricing fair?

For plain scrape/crawl at modest volume, yes — 1 credit per page is transparent and the tiers are competitive with the managed-API category. The fairness erodes in three specific places: the one-time free grant (a surprise, not a scam, but a real friction), the extract dual-billing (the cost most teams underestimate), and the lack of any self-host escape valve to cap the downside. If your workload is pure markdown ingestion at predictable volume, Firecrawl's pricing is livable. If it involves structured extraction, spiky agent traffic, or a hard cost ceiling, model the alternatives before you sign an annual plan.

Sources

• Firecrawl pricing: firecrawl.dev/pricing (verified 2026-05-18)
• fastCRW pricing and self-host: github.com/us/crw · fastcrw.com

Related: Self-host a Firecrawl-like API · Firecrawl vs Crawl4AI vs CRW