Firecrawl
Turns websites into clean, structured data. Feeds our custom research and enrichment steps.
In our stackLast reviewed July 12, 2026
What Firecrawl actually is.
Firecrawl is an API that turns a URL into clean markdown or JSON instead of raw HTML. Point it at one page, a whole site, or a search query, and it hands back content formatted for an LLM to read rather than a browser to render. It runs its own headless-browser layer, so JavaScript-heavy pages and single-page apps get rendered before the content comes back, and it handles PDFs and DOCX files the same way.
The newer Agent mode, still in preview, only ships 5 free runs a day before it moves to its own separate pricing, so it's not something to build a production workflow on yet. Everything else on the platform is credit-based rather than seat-based, and it's open source, so a team that wants to self-host instead of paying per credit can pull the repo directly.
What it does
- Scrapes single pages into markdown, HTML, or structured JSON.
- Crawls a whole site automatically, page after page.
- Also includes Map and Monitor endpoints for site URLs.
- Extracts data behind logins or interactive elements via Interact mode.
- Renders JavaScript-heavy pages and single-page apps automatically, no setup needed.
- Pulls structured fields from pages using a JSON schema you define.
- Reads PDFs and DOCX files the same way it reads web pages.
Most GTM teams here only need Scrape and Crawl; the rest sits there for when the job gets more complicated.
Where it sits in the pipeline.
- AI & Intelligence
- Data & Enrichment
- Scraping & Custom Data
- Email Infrastructure & Deliverability
- Outreach & Sending
- LinkedIn & Multichannel
- Signals & Intent
- Calls & Meetings
- CRM & Pipeline Ops
Runs well with
The tools we'd run next to Firecrawl, and what each pairing covers.
Pricing, without the fiction.
From the Firecrawl pricing page · verified July 12, 2026
- 1,000 credits/month
- Scrape 1,000 pages
- 2 concurrent requests
- Low rate limits
- 5,000 credits/month
- Scrape 5,000 pages
- 5 concurrent requests
- Basic support
- 100,000 credits/month
- Scrape 100,000 pages
- 50 concurrent requests
- Standard support
- 500,000 credits/month
- Scrape 500,000 pages
- 100 concurrent requests
- Priority support
- 1,000,000 credits/month
- Scrape 1,000,000 pages
- 150 concurrent requests
- Priority support
The bill is driven by page-equivalent credits, not seats: Scrape/Crawl/Map cost 1 credit per page, Search costs 2 credits per 10 results, and Interact (headless-browser automation on dynamic pages) costs 2 credits per browser-minute. A workload that leans on Interact for JS-heavy or login-gated sites will burn through credits far faster than a straightforward markdown scrape of static pages, even though both look like "one job" on the surface. Concurrency is also tier-gated (2 concurrent requests on Free up to 150 on Scale), so a large crawl on a cheap plan costs the same in credits but takes much longer in wall-clock time. Standard ($83/mo for 100,000 credits) is Firecrawl's own recommended middle tier; anything past Scale ($599/mo, 1M credits) moves to custom Enterprise pricing.
Choose it, or skip it.
Choose it if
- Dev teams building an AI app (RAG, agent, or research tool) that needs web pages turned into clean markdown or JSON without maintaining their own scraper or headless browser fleet.
- Teams on a usage-based budget who want to start free (1,000 credits/mo) and scale spend with actual page volume rather than buying a flat seat license.
- Builders who need crawl, search, and scrape as one API rather than stitching together separate tools for site discovery, page fetch, and content cleanup.
- GTM/outbound teams that need to pull structured data off prospect or competitor sites (pricing pages, job boards, docs) into a format an LLM can read directly.
Skip it if
- Teams that only need to scrape a handful of static pages occasionally — a simple HTTP fetch or a no-code scraper is cheaper and simpler than standing up an API integration.
- Non-technical users with no one to write API calls; Firecrawl is API/SDK-first with no point-and-click UI for building scrape jobs.
- High-volume crawls of JS-heavy or interaction-heavy sites on a tight budget: Interact mode burns 2 credits per browser-minute, which adds up fast compared to plain page scrapes at 1 credit each.
If not Firecrawl, then what.
- ApifyFree tier ($5/mo usage credit) then paid plans from $29/mo (Starter), usage-based overage billing
Choose Apify when you need a marketplace of pre-built, site-specific scrapers (LinkedIn, Amazon, Google Maps, etc.) rather than a general-purpose fetch-and-clean API.
- Browse AIFree tier (50 credits/mo) + paid plans from $19/mo (annual billing)
Choose Browse AI when the person building the scraper is non-technical and wants a point-and-click recorder instead of writing API calls.
- PhantomBusterfrom ~$56-69/mo (Starter plan), 14-day free trial, a limited free tier (2h execution, 5 slots, capped credits); annual billing gets ~20% off
Choose PhantomBuster when the target is specifically social platforms (LinkedIn, Instagram) with pre-built automations rather than general website content extraction.
Asked in real client work.
What exactly counts as a credit?
Per Firecrawl's pricing page: Scrape, Crawl, and Map each cost 1 credit per page. Search costs 2 credits per 10 results. Interact (browser automation on dynamic pages) costs 2 credits per browser-minute. Monitor costs 1 credit per page check. Credits are pooled across all these features on a plan, so a crawl-heavy month burns down the same pool as a search-heavy one.
How much does it cost at real volume, not the free tier?
Paid plans are billed yearly at these credit/price points: Hobby $16/mo for 5,000 credits, Standard $83/mo for 100,000 credits, Growth $333/mo for 500,000 credits, Scale $599/mo for 1,000,000 credits. Anything beyond that is a custom Enterprise quote with unlimited pages and SLA/SSO/zero-data-retention options.
What happens if I go over my monthly credits?
The pricing page does not publish auto-overage rates on the standard tiers; heavier or unpredictable usage is meant to route to a higher tier or Enterprise custom pricing. Anyone with volume that could spike unpredictably should confirm the overage/rate-limit behavior directly with Firecrawl before committing to a plan.
Does concurrency change with plan tier, and does that matter?
Yes. Concurrent request limits scale by plan: 2 on Free, 5 on Hobby, 50 on Standard, 100 on Growth, 150 on Scale. For a large crawl job, low concurrency on cheaper tiers means the job simply takes longer wall-clock time even though the credit cost per page is the same.
Is the output actually clean enough to feed an LLM without extra processing?
Firecrawl's core pitch is markdown/JSON output built specifically for LLM consumption, and it handles JS-rendered pages via its own browser layer rather than requiring the caller to run headless Chrome. Quality on any given site still depends on that site's structure and anti-bot measures, so spot-checking output on target domains before relying on it in a pipeline is reasonable.