POST /v1/extract

Scrapes the page, then passes the content to Claude with your schema and optional prompt. Returns structured JSON — no regex, no CSS selectors.

Cost: 5 credits per request

Request

curl -X POST https://www.webglean.com/v1/extract \
  -H "Authorization: Bearer wg_your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://news.ycombinator.com",
    "schema": {
      "stories": [{ "title": "string", "points": "number", "url": "string" }]
    },
    "prompt": "Extract the top stories from the page"
  }'

Body parameters

Schema format

The schema is a plain JSON object describing the shape of data you want back. Use primitive types as strings:

{
  "title": "string",
  "price": "number",
  "inStock": "boolean",
  "features": ["string"]
}

Response

{
  "success": true,
  "data": {
    "stories": [
      { "title": "Show HN: I built a web scraper", "points": 342, "url": "https://..." },
      { "title": "Ask HN: Best tools for AI pipelines?", "points": 218, "url": "https://..." }
    ]
  }
}

Use cases

• Extract product prices, inventory, and descriptions from e-commerce pages
• Pull contact info, bios, or job listings from company pages
• Summarize article metadata (author, date, tags, word count)
• Parse event details (name, date, location, ticket price)

Errors