read_page
One web page as clean Markdown, HTML or its links, with JavaScript pages rendered first. ₹0.57 per page; a host that cannot be reached is ₹0.
What it does
The content of one page as clean Markdown (formats: ["markdown"]), raw HTML or its list of links, with the title and metadata, in data.data. Use it to read an article, a pricing page or a product page the assistant cannot open itself. One URL per call: it does not crawl a site and does not search (search_web). A PDF comes back as base64 text.
Inputs
Generated from the contract captured from staging on 2026-11-05.
| Field | Type | Required | Description |
|---|---|---|---|
url | string | yes | e.g. "https://example.com" |
maxAge | integer | no | e.g. 0; min 0, max 172800000 |
formats | array of string (markdown, html, links) | no | e.g. ["markdown"] |
parsers | array of string | yes | e.g. [] |
onlyMainContent | boolean | no | default true |
What drives the cost
parsers must be [] exactly: that keeps the flat price (other parsing options are charged extra at the source and are not offered). onlyMainContent true (the default) strips navigation and footers. maxAge: 0 forces a fresh fetch; a larger value in milliseconds accepts a recent cached copy, up to 48 hours.
Price today
| Price today | ₹0.57 per page (cost_paise: 57) at ₹93.93 per US dollar on 17 November 2026 |
| Hold before the call | ₹0.86 (hold_paise: 86); when the call settles the real price is charged and the unused part is released |
| ₹0 when | it fails, or the provider answers but there is nothing in it (an empty list, no email, no page): class: "ok_free", ₹0 |
| Timeout | 30 s, then upstream_timeout (₹0, retryable) |
| Over MCP | run_tool with tool: "read_page" |
| Over REST | POST https://api.parel.si/v1/run with "tool": "read_page" |
| Confirmation | not needed (the estimate is under the ₹200 threshold) |
A page that answers 403 or 404 is still a fetched document and is charged; a site that cannot be reached at all is ₹0.
Request
curl https://api.parel.si/v1/run \
-H "Authorization: Bearer $PAREL_KEY" \
-H "Content-Type: application/json" \
-d '{"tool":"read_page","input":{"url":"https://example.com/pricing","formats":["markdown"],"parsers":[]}}'
What comes back
data.data.markdown (or html, links) and data.data.metadata with title, sourceURL, statusCode and contentType. A recorded answer from the source, trimmed:
{
"success": true,
"data": {
"markdown": "# Example Domain\n\nThis domain is for use in documentation examples.",
"metadata": {
"title": "Example Domain",
"sourceURL": "https://example.com",
"statusCode": 200,
"contentType": "text/html"
}
}
}
Long pages produce long Markdown. Over MCP strings over the budget are shortened with a _truncated note; REST returns everything.
Notes
- parsers must be [] exactly. formats takes up to three of markdown, html, links.
- A list that spans pages: map_site finds the page URLs in one call, read_pages reads up to ten of them in one call, extract_structured returns the entries as a table. Reading page after page with this tool is the slow way.
- A 403 or 404 page is still a document and is charged; a dead host is free.
- Long pages produce long Markdown; over MCP long strings are shortened, REST returns everything.
- Sites that block bots return a short interstitial as the page: check the content before trusting it.
Siblings
search_webto find the page first.search_newswhen the page is an article you have not found yet.
This page as Markdown: /docs/tools/read_page.md. Something wrong or missing: hello@parel.si.