The web access layer
for AI agents
Any URL to clean, LLM-ready markdown — rendered in a real browser. REST and MCP, one flat credit per format.
# Example Domain This domain is for use in documentation examples without needing permission. Avoid use in operations. [Learn more](https://iana.org/domains/example)
Ready · Live — a real browser runs it
Features
Everything a page needs to become data — one flat credit per format.
JavaScript rendering
Every page renders in a real browser — SPAs included, no empty shells.
Clean markdown
Structured, de-cluttered output built for LLMs — not raw HTML.
MCP native
Twelve tools for Claude, Cursor, Codex and friends.
Monitors
Scheduled re-scrapes with change detection, included on every plan.
Structured extraction
Schema-driven JSON from any page — products, listings, articles.
Cached by default
Repeat scrapes return in milliseconds; maxAge controls freshness.
Compare
How fastcrawl stacks up.
| Feature | fastcrawl | Firecrawl | Jina Reader | Tavily | Exa |
|---|---|---|---|---|---|
| Free tier | 1,500 credits / mo | 1,000 credits / mo | 10M tokens | 1,000 credits / mo | $10 credit / mo |
| Paid entry | $5 / mo | $19 / mo (billed yearly) | $50 top-up | $30 / mo | pay-as-you-go |
| Credits at entry plan | 5,000 / mo | 5,000 / mo | token-metered | 4,000 / mo | usage-metered |
| Rate limit on paid | None — metered overage | 25 concurrent | — | — | 10 QPS |
| One flat credit, every format | ✓ | — | — | — | — |
| JS rendering, SPAs included | ✓ | ✓ | ✓ | ✓ | ✓ |
| Site-wide crawl | ✓ | ✓ | — | ✓ | — |
| Screenshot + PDF capture | ✓ | ✓ | — | — | — |
| LLM structured extract | 1 credit | extra credits | — | credits | AI summaries extra |
| Scheduled monitors + change alerts | ✓ | ✓ | — | — | ✓ |
| MCP server, zero config | 12 tools | ✓ | ✓ | ✓ | ✓ |
| Web search endpoint | ✓ | ✓ | ✓ | ✓ | ✓ |
| Neural / semantic search | — | — | — | — | ✓ |
| Median fresh scrape | 325ms HTTP / ~6s browser | — | ~7.9s* | — | 180ms–1s (search) |
Competitor figures from their public pricing pages and docs, verified 29 August 2026. Firecrawl's $19 Hobby tier is the yearly-billed price. * Jina's own published average latency. Exa's $10/mo is a recurring credit grant, not a request count.
Endpoints
Twelve endpoints. One key.
scrape · crawl · map · search · extract · screenshot · pdf · parse · batch · monitors · usage · me Full reference →
Pricing
Start free. Scale when you're ready.
Free
$0 / month
For trying the API and small agents.
- 1,500 credits monthly
- 100 scrapes per day
- 20 requests per minute
- All endpoints, one credit each
- No card required
Go
$5 / month
Best value — production-ready agents.
- 5,000 credits monthly
- No rate limit — pay-as-you-go overage
- Webhooks & monitors
- Email support
Enterprise
Custom
High volume, custom limits.
- Custom credits & limits
- SLA & SSO
- Dedicated account manager
- Volume pricing
Every format costs one flat credit · monitors included on every plan · 1.5× the free credits of Firecrawl, from $5/mo
Overage on Go: $1 per 1,000 extra credits · need a custom limit? Contact us
FAQ
Questions, answered.
How is this faster and cheaper than Firecrawl or Jina?
Fresh scrapes come back in about 325ms at the median on the HTTP path, and about 6 seconds when a real browser render is needed—both faster than Jina's ~7.9s average. Free tier gives 1,500 credits—1.5× Firecrawl's—and Go starts at $5, under a third of their $19 floor.
How is this different from running Playwright myself?
You'd need a server, a browser fleet, session handling, retries, anti-bot workarounds and scaling — for every page. Fastcrawl runs Cloudflare's edge browsers for you and returns only the markdown. One API call instead of an infrastructure project.
Does it work on JavaScript-heavy sites?
Yes. Every page renders in a real browser before extraction — SPAs included. If a render comes back thin, we detect it and retry through the patient path automatically, so you get the full content or a clear error — never a broken shell.
What counts as a credit?
One credit buys one unit of work, regardless of format: a scrape, a crawl page, a search, a screenshot, a PDF, or one monitor run — no per-endpoint surcharges. Cached repeats are free.
How fresh is the data?
By default, repeat scrapes of the same URL within 48 hours are served from cache — milliseconds, not seconds. Set maxAge=0 to always render fresh, or anything in between.
Give your agents the whole web.
Sign up in thirty seconds, copy your key, and point your agent at any URL.
Start free — no card