Scrape

POST /v1/scrape fetches one URL and converts it. The model is <provider>/<model>. The provider fetches the page, not Aile, and nothing is stored.

1curl https://api.aile.sh/v1/scrape -H "Authorization: Bearer $AILE_KEY" -H "Content-Type: application/json" -d '{ "model": "firecrawl/scrape", "url": "https://example.com", "formats": ["markdown", "links"] }'

1{
2  "object": "scrape",
3  "model": "firecrawl/scrape",
4  "url": "https://example.com/",
5  "markdown": "# Example Domain ...",
6  "links": ["https://www.iana.org/domains/example"],
7  "metadata": { "title": "Example Domain", "status_code": 200, "content_type": "text/html" }
8}

Models

A field a model does not take is a 400, never silently dropped.

ScrapingDog and ScrapingBee answer markdown or raw_html with an empty metadata, and return HTML pages only: a PDF or other binary is a 415. ScrapingBee refuses Google URLs.

Fields

• url: required, http or https.
• formats: default ["markdown"]. ScrapingDog and ScrapingBee take exactly one.
• render: static or js (runs the page's JavaScript, at a higher price). Absent means static.
• wait_for: milliseconds, up to 10,000. On ScrapingDog and ScrapingBee it needs render: "js".
• country: two-letter code. Optional on Firecrawl, required on scrapingdog/scrape-geo, refused on scrapingdog/scrape and scrapingbee/scrape.
• only_main_content: Firecrawl, default true. ScrapingDog returns the whole page; on ScrapingBee markdown is the main content and raw_html the whole page.
• mobile: Firecrawl only.
• max_pdf_pages: Firecrawl, 1 to 50, default 50; caps how much of a PDF is read. The other models take at most 1.

Billing

• Firecrawl: a web page is one unit, a PDF one unit plus one per page read. The hold is 1 + max_pdf_pages and is written down to what was read. Lower max_pdf_pages for a smaller hold.
• ScrapingDog and ScrapingBee: one unit per request at the price of the render you named, so the hold is the bill.
• ScrapingDog and ScrapingBee: a target page that answers 404 or 410 is a 404 and costs nothing (413 is a 413). Firecrawl returns the page with its metadata.status_code and bills it.
• Keyless x402 works here when the deployment enables it. You pay for what you name: on Firecrawl, with no max_pdf_pages, one webpage and no PDF parsing; with it, 1 + max_pdf_pages webpages, whatever the page turns out to be. On the other models, one webpage at the render price.