read_pages

Up to ten web pages as clean text, highlights or a summary in one call. ₹0.39 per page per content type; a page that cannot be fetched is ₹0.

What it does

The content of up to ten urls in one call, in data.results[] with title, url, author, publishedDate and what you asked for: text (clean page text, {maxCharacters: N} to cap it), highlights ({query, numSentences} for the passages most relevant to a question) or summary ({query}); data.statuses[] says which URLs failed. Use it to read several pages at once; for one page with JavaScript rendering and Markdown use read_page.

India: no locale knob.

Inputs

Generated from the contract captured from staging on 2026-11-05. This tool is priced per page: the hold follows the number of urls you send.

FieldTypeRequiredDescription
textbooleanobjectnotrue, or {maxCharacters: N} — the page as clean text. e.g. {"maxCharacters":2000}
urlsarray of stringyesone or more page URLs; each page is billed per content type requested. e.g. ["https://en.wikipedia.org/wiki/Hierarchical_navigable_small_world"]
summaryobjectno{query} — an AI summary of the page focused on query
highlightsbooleanobjectnotrue, or {query, maxCharacters} — the passages most relevant to query
maxAgeHoursintegerno

What drives the cost

The number of URLs times the content types you ask for. The hold covers urls × all three types; the charge settles at what you asked for (text alone is a third). More than ten URLs is invalid_input, ₹0. A URL that cannot be fetched appears in statuses[] and is not charged.

Price today

Price today₹0.39 per page (cost_paise: 39) at ₹93.93 per US dollar on 17 November 2026; each one asked for is charged, ₹3.39 for the largest call
Hold before the call₹0.59 (hold_paise: 59) for one, ₹5.09 for the largest call; when the call settles the real price is charged and the unused part is released
₹0 whenit fails, or the provider answers but there is nothing in it (an empty list, no email, no page): class: "ok_free", ₹0
Timeout30 s, then upstream_timeout (₹0, retryable)
Over MCPrun_tool with tool: "read_pages"
Over RESTPOST https://api.parel.si/v1/run with "tool": "read_pages"
Confirmationnot needed (the estimate is under the ₹200 threshold)
Count fieldurls: The hold covers the number of results you ask for in urls; leave it out and the worst case is held.

Request

curl https://api.parel.si/v1/run \
  -H "Authorization: Bearer $PAREL_KEY" \
  -H "Content-Type: application/json" \
  -d '{"tool":"read_pages","input":{"text":{"maxCharacters":2000},"urls":["https://en.wikipedia.org/wiki/Zomato"]}}'

What comes back

data.results[] and data.statuses[]. A real answer recorded during the live check of 2026-10-28, trimmed:

{
  "results": [
    {
      "id": "https://en.wikipedia.org/wiki/Zomato",
      "title": "Zomato",
      "url": "https://en.wikipedia.org/wiki/Zomato",
      "author": null,
      "text": "Zomato\n\nZomato\n\n| | |\n| --- | --- |\n| Former name | Foodiebay (2008–2018) |\n| Website type | Online food ordering |\n| Founded | 10 July 2008; 18 years ago [1] |\n| Headquarters | Gurugram, Haryana, India |\n| Area served | India |\n| Owner | Eternal Limited |\n| Founders | Deepinder Goyal Pankaj Chaddah |\n| Key people | Deepinder Goyal (CEO & MD) |\n| Services | …"
    }
  ],
  "statuses": [
    {
      "id": "https://en.wikipedia.org/wiki/Zomato",
      "status": "success",
      "source": "cached"
    }
  ],
  "searchTime": 16.83979699993506
}

Notes

  • Up to 10 urls per call (more is invalid_input, free). Each page is charged once per content type you ask for; text alone is cheapest.
  • maxAgeHours accepts a recent cached copy; omit it for the source's default freshness.
  • A URL that cannot be fetched appears in statuses[] with an error and is not charged.

Siblings

This page as Markdown: /docs/tools/read_pages.md. Something wrong or missing: hello@parel.si.