fastCRW Pricing — Search, Scrape, Crawl, Map in One Plan
Start building in minutes, or deploy on your own servers.
Managed infrastructure
Five tiers, one continuous ladder. Standard is where most production agents land — but the floor is always $0 self-hosted.
Enterprise
Beyond 1M credits/mo — dedicated clusters, premier support.
Scale
$5,604 billed yearly · save $984/yr
$0.55 / 1k credits
For teams scaling their data pipelines.
Growth
$2,844 billed yearly · save $504/yr
$0.56 / 1k credits
Built for high volume and speed.
Standard
$708 billed yearly · save $120/yr
$0.69 / 1k credits
Perfect for scaling with less effort.
Unit cost falls as you scale: $2.60 / 1k on Hobby base → $9.00 at Hobby overage → $0.69 / 1k on Standard. Around 11,222 pages/mo Standard's flat rate overtakes Hobby + overage — the higher tier is spike insurance, not a penalty.
Why does the in-product upgrade wall pre-select Hobby while this grid defaults to Standard? The wall pre-selects Hobby because the user just proved 500-credit-scale usage, not Standard-scale — the pricing grid's Standard-default is the cold-anchor; the wall's Hobby is the warmed-anchor; both are intentional.
Concurrency figures show each plan's provisioned target. During the current capacity rollout (as of 2026-05-17), sustained concurrent throughput may be lower than the figure shown while infrastructure scaling completes; requests over live capacity receive a 503 with a Retry-After header. See Rate Limits for retry guidance.
No lock-in — a legal right, not a vendor promise
Your exit isn't a vendor promise — it's a legal right. The engine is AGPL-3.0: the software you depend on is already on your disk under a license no one can revoke. Dislike our Cloud or our price → self-host the same API, no migration.
What the same workload actually costs
Pages/month, billed as pages (1 credit = 1 page on plain scrape/crawl for both). Firecrawl figures are locked & dated [verified: 2026-05-18] (Standard shown at the charitable annual-effective $83, not the higher $99 monthly-list). Self-host is the $0 floor: the price is zero, you bring the infra.
| Pages / month | Firecrawl | fastCRW (best tier) | Self-host (AGPL) |
|---|---|---|---|
10,000 Small project | $83/mo | $69/mo Standard | $0 one-time / lifetime infra TCO floor |
11,222 Hobby→Standard crossover Past this volume Standard's flat rate beats Hobby + overage — spike insurance, not a cliff. | $83/mo | $69/mo Standard | $0 one-time / lifetime infra TCO floor |
100,000 Production agent | $83/mo | $69/mo Standard | $0 one-time / lifetime infra TCO floor |
500,000 High volume | $333/mo | $279/mo Growth | $0 one-time / lifetime infra TCO floor |
fastCRW cost = plan base + block-billed overage, computed directly from the live price list. At the 11,222 pages/mo crossover the flat Standard rate overtakes Hobby + overage — the higher tier is spike insurance, not a penalty.
Cloud or self-hosted. Your call.
Managed Cloud
500 free credits to start, no card required
Included monthly allowance on paid plans
Manual top-up packs that carry until consumed
Optional auto-recharge to avoid interruptions
Forever Free
AGPL-3.0 licensed — no license cost ever
Unlimited requests and concurrency
One small, self-contained binary
Full control over data and infrastructure
Enterprise Volume
Frequently asked questions
Are the 500 free credits monthly?
No — one-time for the lifetime of the account, not a monthly reset. Paid plans have monthly allowances.
Does using fastCRW Cloud put my code under AGPL?
Using the fastCRW Cloud API places no open-source obligation on your code — AGPL applies only if you self-host and distribute a modified engine. Need private self-hosting without copyleft? A commercial license is available.
How do credits work?
Each paid plan starts each month with its full included credit allowance. Purchased top-up credits carry over until consumed. Optional auto-recharge keeps jobs running without surprise overages.
When should I choose cloud instead of self-hosting?
Choose cloud when you want immediate access, hosted scaling, and minimal operational ownership. Choose self-hosting when cost control, privacy, or infra ownership matters more.
Is self-hosting really free?
Yes. CRW is AGPL-3.0 licensed. Deploy on your own servers with unlimited requests and zero license cost. You only pay for your own infrastructure.
What happens when I run out of credits?
API requests return a 429 status code. You can buy a top-up pack instantly from the billing dashboard or enable auto-recharge to avoid interruptions.
When does launch pricing end?
Launch pricing ends June 1, 2026. If you subscribe before that date, your current rate is locked for 12 months after launch ends. No surprise price jumps.
How is /v1/extract billed?
Token-based dynamic pricing: you pay 3× the provider's LLM cost converted to credits (1 credit = $0.001 USD). Cached input tokens are discounted at the provider's cache-hit rate. Each request reserves a worst-case credit estimate up front and is reconciled to the actual token count after the LLM call returns — refunds for the unused portion land in the same transaction as the charge.
Can I bring my own LLM key on /v1/extract (BYOK)?
Yes. Pass 'llmApiKey' in the request body and fastCRW charges a flat 1-credit infra fee instead of marked-up token pricing. BYOK works on every plan, including FREE, and the key is never persisted or logged.
Is /v1/extract available on the FREE plan?
Managed /v1/extract (we pay the LLM) requires HOBBY or higher. FREE accounts can still use /v1/extract via BYOK — supply your own LLM key and pay only the 1-credit infra fee. /v1/scrape, /v1/crawl, /v1/map, and /v1/search itself work on every plan; the same managed-vs-BYOK split applies to /v1/search's answer/summarizeResults flags (see the next two FAQs).
Does /v1/search support managed answer/summarize?
Yes. answer:true and summarizeResults:true now work without llmApiKey on HOBBY and above. We use Anthropic Claude Sonnet 4 by default and bill token-by-token against the provider price table with a 3× markup. FREE accounts still need to bring their own key on those flags — pass llmApiKey to bypass managed pricing on any plan.
What does managed /v1/search cost per call?
We pre-reserve a worst-case credit estimate (capped at 8 000 credits per request), then reconcile to the actual cost from the engine's llmUsage report. Successful summaries and the answer leg are summed across one response; per-leg max_tokens is capped at 1024 so the actual charge stays within the reservation. The Decision-11 dispatch surfaces zero-cost, no-op, and fail-closed branches in the response _meta field so you can see exactly which path billed.
What changed about scrapeOptions.formats:"summary" on /v1/search?
scrapeOptions.formats:"summary" was the BYOK-only escape on /v1/search. It is removed — use summarizeResults:true instead, with or without llmApiKey. /v1/scrape still accepts formats:"summary" but now requires llmApiKey; without one, use /v1/extract (managed available on HOBBY+) or pass the BYOK key on the call.
Who builds fastCRW?
fastCRW is built by an independent engineering team focused on making web data extraction fast, reliable, and affordable. The core engine (CRW) is open-source under AGPL-3.0.
Compare and explore further
See how fastCRW stacks up against other tools: Firecrawl alternative, Tavily alternative, Exa alternative. Or browse all alternatives and benchmarks. Building an agent? See the AI agents use case and MCP integration. Full reference at docs.fastcrw.com.