Home
DESIGN.mdStar on GitHub

Power AI agents with
clean web data

The context API to search, scrape, and interact with the web at scale. Turn any website into LLM-ready markdown or structured data.

$ pip install firecrawl-py
Colors
Brand
Surfaces (light)
Text
Illustrative flame & scrape state
Typography
Inter / 600
≈ Suisse · Display · 52px
Power AI agents with clean web data
Inter
Inter / 600
Heading · 20px
Search, scrape, and crawl the whole web
Inter
Inter / 400
Body · 15px
One API turns any URL into clean, LLM-ready data.
Inter
Geist Mono
Code / data · 13px
app.scrape_url('https://firecrawl.dev')
Geist Mono
Signature · The scrape
POST /v1/scrape Stripe Docs Ready
Target URL
GET
Preset
Output format
Pipeline
response · markdown
Raw web → clean data. Pick a page, choose a format, and Scrape. Firecrawl strips the nav, ads, and cruft — returning just the content as Markdown, structured JSON, or a link map.
Components
Buttons
Scrape state
● 200 Scraped ● Crawling ● Rate-limited ● 404 Failed
/scrape /crawl /search /extract
API endpoint (input)
scrape → crawl → extract
Radius
btn · 7
field · 9
panel · 14
Guardrails
Do
Keep the canvas near-white #FDFDFD with crisp black ink; let one heat-100 orange carry the brand.
Set every endpoint, JSON key, and extracted line in Geist Mono — code and data are always mono.
Show the core value: raw web → clean markdown. The output panel is the point.
Fill the flame mark with heat-100, always.
Don't
Use the flame gradient or agent-purple as the primary UI accent — they're illustrative only.
Add a second competing bright color — green/yellow/red are scrape-state signals only.
Render extracted content in a proportional sans — it's data; keep it mono.
Recreate the flame in another hue.