Web Page Data Extraction

Example

{
  "url": "https://example.com"
}

{
  "data": {
    "url": "https://example.com",
    "title": "Example Domain",
    "metadata": {
      "page_type": "informational"
    },
    "main_content": {
      "links": [
        {
          "url": null,
          "text": "Learn more"
        }
      ],
      "headings": [
        "Example Domain",
        "Example Domain"
      ],
      "body_text": "This domain is for use in documentation examples without needing permission. Avoid use in operations."
    }
  },
  "page_title": "Example Domain",
  "source_url": "https://example.com"
}

Inputs

Output fields

How reliably each field appears in a successful answer.

How to call it

The same tool through each route. MCP clients reach it through Strale’s server at https://api.strale.io/mcp; the call below is what the agent sends.

# Needs an API key. max_price_cents caps what this call may cost.
curl -X POST https://api.strale.io/v1/do \
  -H "Authorization: Bearer $STRALE_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "capability_slug": "web-extract",
    "inputs": {
      "url": "https://example.com"
    },
    "max_price_cents": 15
  }'

{
  "tool": "strale_execute",
  "arguments": {
    "slug": "web-extract",
    "inputs": {
      "url": "https://example.com"
    },
    "max_price_cents": 15
  }
}

import { Strale } from "straleio";

const strale = new Strale({ apiKey: process.env.STRALE_API_KEY! });
const result = await strale.do({
  capability_slug: "web-extract",
  inputs: {
    "url": "https://example.com"
  },
  max_price_cents: 15,
});

import os
from straleio import Strale, DoRequest

strale = Strale(api_key=os.environ["STRALE_API_KEY"])
result = strale.do(DoRequest(
    capability_slug="web-extract",
    inputs={
        "url": "https://example.com",
    },
    max_price_cents=15,
))

# No account. The first request answers 402 with the price in USDC;
# an x402 client pays it from your wallet and repeats the request.
curl -i -X POST https://api.strale.io/x402/v2/web-extract \
  -H 'Content-Type: application/json' \
  -d '{
    "url": "https://example.com"
  }'

Limitations

• Some website sources are unsupported

  Trustpilot, Glassdoor, LinkedIn, Google Search, Google Patents, Reddit and the major social platforms (Facebook, Instagram, X) are not supported and return a policy error before the page is read. Repeating the same request does not return content.

• Login, CAPTCHA and multi-step pages are inaccessible

  Renders pages via headless browser — pages requiring login, CAPTCHA, or multi-step navigation cannot be accessed

  What to do instead: Use an accessible public page containing the information you need.

• Complex layouts can lose structural context

  AI extraction interprets page content — complex data tables or deeply nested layouts may lose structural context

  What to do instead: Use structured-scrape with CSS selectors when you know the page structure, and review complex tables against the original.

• Output size cap per call

  Structured output is limited to about 16000 output tokens per call; requests exceeding this limit are refused rather than returned truncated. Requests with large results can exceed this limit, and repeating the same request does not help.

  What to do instead: Narrow the 'extract' instruction, or split the request into multiple calls (e.g. one page or one section at a time)

• Pages are limited to 16 MiB of HTML

  Pages with more than 16 MiB of HTML are refused with a caller_input error naming the input field. The call is not charged, and the page is not retried through another fetch or rendering route.