Spicrawl now works with OpenAI Agents SDK, LangChain, Codex, Claude Code and more.See integrations
Skip to content

Extract structured data from any website

Pick the fields you need. Get them back as clean JSON.

example.com/product/1
Meet Model X — Precision in every click
Wireless Mouse — Model X
Wireless Mouse — Model X
$49.00In stock★ 4.6 (1,284)
Low-latency 2.4 GHz receiver, 70-day battery life and silent switches.
SpecValue
Connection2.4 GHz, BT5Battery70 days
<script type="application/ld+json">
result.md
# Wireless Mouse — Model X**$49.00** · In stock · 4.6 (1,284 reviews)Low-latency 2.4 GHz receiver, 70-day battery lifeand silent switches.| Spec      | Value        || --------- | ------------ || Connection| 2.4 GHz, BT5 || Battery   | 70 days      |
Reading the page
Raw HTML12,480 tokens
Markdown0 tokens

Menus, footers and ads removed: 3% to 24% of the raw HTML tokens. Example page.

How it works

  1. Name your fields and give each one a CSS selector.
  2. Send them in extract with the page URL.
  3. Read the JSON. Your fields are under data.
curl https://api.spicrawl.com/v1/scrape \
  -H "Authorization: Bearer $SPICRAWL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/products/42",
    "extract": { "title": "h1", "price": ".price", "next_page": "a.next@href" }
  }'
json
{
  "status": 200,
  "data": {
    "title": "Walnut Desk",
    "price": "$349.00"
  },
  "empty_fields": ["next_page"]
}

empty_fields lists fields that matched nothing. It is often the first sign that a site changed its layout.

Three ways to get data

WayWhat you sendWhat you get
autoparse"autoparse": trueThe page’s own JSON-LD, OpenGraph, microdata, metadata and app state
Selector mapField names and CSS selectorsExactly those fields, as text
JSON SchemaA schema whose fields have selectorsTyped, checked values, so "$1,234.56" becomes 1234.56

Good to know

  • No extra credits for extraction. You pay only for the request: 1 credit plain, 3 with JavaScript.
  • Coming soon: extraction from a plain-language prompt. Today, use selectors, a schema or autoparse.
  • Not on batch. Extraction runs on single scrapes only.
  • Your responsibility: only extract data you have the right to use. See our terms.

Frequently asked questions

How do I extract data from a website as JSON?

Send extract with a map of field names to CSS selectors, such as "title": "h1" and "price": ".price". The values come back under data in the JSON response.

What does autoparse do?

It returns what the page already says about itself, with no selectors: JSON-LD, OpenGraph, Twitter Card, microdata, RDFa, head metadata and embedded app state.

Can it return numbers instead of text?

Yes. Use a JSON Schema whose properties have a selector. Values are converted to the declared type, so "$1,234.56" becomes 1234.56, and the result is checked against the schema.

How do I know when a site changes its layout?

Check empty_fields in the response. It lists the fields whose selector matched nothing, which is often the first sign that the markup changed.

Can I describe the fields in plain words instead of selectors?

Not yet. Extraction with a model prompt is coming soon. Today, use selectors, a JSON Schema or autoparse.