Quick start

Create a key in the dashboard, then:

curl "https://stealthasf.bunnybag.space/v1/scrape" \
  -H "x-api-key: sasf_live_…" \
  -H "content-type: application/json" \
  -d '{"url": "https://quotes.toscrape.com/", "engine": "auto", "extract": "links"}'

Authenticate with x-api-key: <key> or Authorization: Bearer <key>. Browser requests can take up to ~2 minutes — set your client timeout to 180 s.

Request — POST /v1/scrape

field	type	meaning
url	string, required	http(s) URL, up to 2048 chars. Internal/private addresses are refused.
engine	auto · http · browser · stealth	auto (default) starts on plain HTTP and steps up to the stealth browser only when the site answers with an anti-bot wall; it also remembers sites that needed stealth. http: no browser. browser: real browser for JavaScript-built pages. stealth: anti-detect browser that clears Cloudflare Turnstile.
extract	links · text · meta · emails · images · headings · table	Return structured data instead of parsing HTML yourself.
steps	array, up to 40	Browser actions run in order on the page (see below). Forces a browser engine.
randomize	{delays, motion, order}	Human-like pauses, mouse and scroll between steps.
repeat, repeat_gap_ms	int, int (≤ 60000)	Run the steps N times in one browser session with a pause between (e.g. read a price every 10 s). Stops after ~14 min.

Browser steps

Each step is a one-key object. Selectors start with css: or xpath:; append ::text or ::attr(href) to read text or an attribute.

step	example
crawl	{"crawl": {"links": "css:a.item", "next_page": "css:a.next", "fields": {"title": "css:h1::text"}, "max_items": 500}} — opens every detail link on each list page and extracts fields there, following next_page. One Cloudflare clearance for the whole crawl. Without links, extracts item_selector rows from the list pages.
scrape	{"scrape": {"item_selector": "css:.row", "fields": {"name": "css:.name::text"}}}
click · hover · scroll · wait_for	{"click": "css:button.more"}
type	{"type": {"selector": "css:input[name=q]", "text": "boots"}}
press_key	{"press_key": "Enter"}
goto · wait	{"goto": "https://site/page/2"} · {"wait": 1500}
find_and_click	{"find_and_click": {"text": "Load more"}} — scrolls until the text appears, then clicks.
collect_links	{"collect_links": {"selector": "css:a.result"}}
capture	{"capture": {"name": "price", "selector": "css:.price"}} — stores a value; use it later as {{price}}.
js_eval	{"js_eval": {"name": "items", "code": "return (await fetch('/api/items')).json()"}} — runs in the cleared page with its cookies.
if_exists · try	{"if_exists": {"selector": "css:.cookie-ok", "then": [{"click": "css:.cookie-ok"}]}} · {"try": {"steps": [...]}} (failures inside are ignored)

Response — 200

{
  "ok": true,
  "job_id": "cm…",
  "engine": "stealth",            // the tier that produced the result
  "status": 200,                  // the target's HTTP status
  "html": "<!doctype html>…",
  "data": { "kind": "table", "columns": [...], "rows": [[...]] },   // with "extract"
  "records": [ {"type": "crawl", "url": "…", "data": {...}} ],      // with "steps"
  "discovered_api": { "kind": "json_api", "url": "…", "method": "GET" },  // when the page called its own data API
  "usage": { "engine": "stealth", "bytes": 3145728, "proxy_mb": 3.1, "ms": 53210 },
  "credits_charged": 45,
  "balance": 249955
}

Errors

status	meaning	charged
400	Invalid body, URL, engine, extract or step	no
401	Missing or invalid API key	no
402	Not enough credits (balance included)	no
422	The site's anti-bot blocked the request (job_id, balance included)	no
429	Rate limit (120 requests/min per IP), or a free-plan account already has a request running	no
502	The target could not be reached, or an engine error	no

Credits

what	credits
http request	1 (+1 per 2 MB of response)
browser request	5 (+10 per MB of transfer beyond the first MB)
stealth request	25 (+10 per MB of transfer beyond the first MB)
solved captcha	+25
blocked request	0

With engine: auto you pay only for the tier that succeeded. Images, video and fonts are not downloaded unless you ask for screenshots, so transfer stays small.