README.md
★ Check OSS self-host version
from the docs You know the library. This is the same crawler, hosted. No browsers, no proxies, no cookies to babysit: one key with $10 of free credit, and it works in Claude Code, Codex, OpenCode and Cursor. Always open source, forever.
NEW Finally, your favorite open-source crawler is now a hosted API.Your open-source crawler, now hosted.
Crawl4AI now available on Cloud
Turn any URL into clean Markdown, typed JSON, or search results. Fast, affordable, and efficient - no infra, no signup.
$ crawl scrape
--instruction
Live · play limits✓ 200 OK
# Type a URL above and hit Run - real output renders here
## What you get #
- /scrape URL → clean Markdown, boilerplate stripped, ready for a prompt.
- /extract URL + a plain-English ask → typed JSON records. No selectors.
- /search query → ranked results as JSON - add
rich=1for follow-ups, related & more. - /answer a question → a direct answer with its sources. Experimental.
- at scale queues, jobs, retries, and solid proxies - handled for you.
## Use it in Claude Code, Cursor & agents - paste one line #
Same key everywhere. Mint one above and it drops into each line below.
claude mcp add --transport http crawl4ai https://api.crawl4ai.com/mcp --header "Authorization: Bearer sk_live_…"~/.codex/config.toml
[mcp_servers.crawl4ai] url = "https://api.crawl4ai.com/mcp" http_headers = { Authorization = "Bearer sk_live_…" }opencode.json
"mcp": { "crawl4ai": { "type": "remote", "url": "https://api.crawl4ai.com/mcp", "headers": { "Authorization": "Bearer sk_live_…" } } }~/.cursor/mcp.json · Windsurf: ~/.codeium/windsurf/mcp_config.json
"crawl4ai": { "url": "https://api.crawl4ai.com/mcp", "headers": { "Authorization": "Bearer sk_live_…" } }~/.openclaw/openclaw.json, the mcp.servers block
mcp: { servers: { crawl4ai: { url: "https://api.crawl4ai.com/mcp", transport: "streamable-http", headers: { Authorization: "Bearer sk_live_…" } } } }Claude Desktop, ChatGPT, Gemini CLI, VS Code, n8n, or your own client · streamable HTTP
url https://api.crawl4ai.com/mcp · header Authorization: Bearer sk_live_…export CRAWL4AI_URL=https://api.crawl4ai.com CRAWL4AI_KEY=sk_live_…curl -s "https://api.crawl4ai.com/search?q=web+scraping" -H "Authorization: Bearer sk_live_…"GET https://api.crawl4ai.com/llms.txtOpen## Self-host, or skip the setup #
// run it yourself
Self-host
It's Apache-2.0. pip install crawl4ai, point it at your own boxes and proxies. No gate, no token. It's yours.
// let us run it
Hosted
Or forget the infra. We've operated this crawler at scale for years - proxies, queues, jobs, the hard parts. You keep the key; we keep it fast.
Who runs it better than the people who built it? Self-host for control; host with us to move fast. Same crawler either way.
## Pricing #
Your favourite open-source crawler, now in the cloud. Credits, not seats: you pay very little and get more than anywhere else.
Try it now: $1 of credit for 24 hours, no signup.
tier
price
for
Community
$10 free to start · no card · +$5 every month
Prototypes and side projects. Get a key
Supporter
$10 · $25 · $100 credit packs
Pay as you go. 10 at a time, no monthly cap, credit never expires.
Developer
Coming soon
The subscription: production limits.
Enterprise
Let's talk
Volume, regions, SLA. Contact us
1 credit = $0.001 · page 1 · search ½ · from our archive ½ · real browser 2 · hard site 4
Note
Launch pricing for the first year: discounted, and it may change. Credit you already hold keeps its value. Launch gift, until 31 December 2026: your first $10 pack is on us.Launch gift, until 31 December 2026: $10 of credit to start.## Philosophy #
# we don't hide.
every result carries where it came from - the origin travels with the data.
# and the whole "block the bots" arms race? a phase.
let crawlers crawl.
every result carries where it came from - the origin travels with the data.
# and the whole "block the bots" arms race? a phase.
let crawlers crawl.
