Send a web page URL and tell Scrape which information you need. It reads the page, including content loaded with JavaScript, and returns that information in the fields you requested. Your application can use the results without having to extract them from the page itself.
curl https://api.cloud.gomry.com/v1/scrape \
-H "Authorization: Bearer $GOMRY_CLOUD_KEY" \
-H "Content-Type: application/json" \
-d '{
"url": "https://example.com/products/desk-lamp",
"prompt": "Extract the product name.",
"schema": {"type": "object", "properties": {
"name": {"type": "string"}
}}
}'Use POST /v1/scrape when your application needs specific fields rather than a whole page. You send the URL, instructions and a JSON schema that defines the fields and their data types. For example, a product catalog could extract a published name and description into the same schema across supplier sites, then pass those records to its review queue.
Use POST /v1/fetch when the useful information is already in JSON-LD, an Open Graph tag or another structure your code can parse. You receive the rendered page without paying for LLM extraction, at a separate, lower rate.
Your schema defines the output shape; the prompt tells Scrape what to look for. A product page containing the text below can become a record with the two fields your catalog needs.
Adjustable desk lamp. A desk lamp with an adjustable arm and warm-white LED light.
{
"url": "https://example.com/products/desk-lamp",
"prompt": "Extract the product name and description.",
"schema": {
"type": "object",
"properties": {
"name": {
"type": "string"
},
"description": {
"type": "string"
}
}
}
}{
"name": "Adjustable desk lamp",
"description": "A desk lamp with an adjustable arm and warm-white LED light."
}The extracted values are returned in data. Your application can store the record in its catalog or pass it to a review queue. You can also request rawHtml when a downstream step needs the original content. For Fetch, the request only needs the URL; your parser handles the returned HTML and metadata.
Scrape works with URLs you already have. It does not discover pages or follow links as a crawler. Your application decides which URLs to request next.
JavaScript rendering helps with pages that build content in the browser. For content that arrives later, you can add a wait before reading it. An optional retry can fetch the page again when a required field is missing; both page-returning attempts are billable.
| Operation | Price | Unit |
|---|---|---|
| Scrape extraction | $0.02 | Per page returned |
| Fetch HTML and metadata | 0.5¢ | Per fetch returned |
The response reports pages or fetches so your application can record usage with the result. A timeout, unreachable host or failed request that returns no page is not billed.
Enable Scrape for your project and grant the API key the permissions for the operation you need. Extraction and Fetch have separate permissions. Send the key as a Bearer token with your JSON request.