Skip to content
Skryp

Typed JSON from any page, checked against the page

Extract returns the fields you ask for as JSON. It reads the page's own data first, then selectors learned from earlier pages with the same layout, and only then a small model. Every value is checked against the page text; anything the page does not contain is flagged as unverified instead of being passed off as fact.

  • JSON Schema or a plain-language prompt.
  • Learned selectors make the next page with the same layout free of model cost.
  • Unverified values are listed explicitly.
  • Escalates to a browser when required fields are missing from the plain page.

Limits

  • Very long pages are focused on the requested fields before the model reads them.

Price

The page read, plus 5 credits when the model runs. Page data and learned selectors add nothing. Full rate card

Start freeRead the quickstart
Extract a price · REST
curl -X POST https://api.skryp.dev/v1/extract \
  -H "Authorization: Bearer $SKRYP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "url": "https://shop.example/kettle",
  "prompt": "product name, price, currency, in stock"
}'
Extract a price · MCP tool call
{
  "tool": "extract",
  "arguments": {
    "url": "https://shop.example/kettle",
    "prompt": "product name, price, currency, in stock"
  }
}

Related