Skip to the page
Get started free

Extract any page as markdown or text

Send up to 20 URLs in one call and get the main text of each, without its menus, footers, and ads. A page that refuses a plain read is tried again, and read_from says where the text came from.

Get an API keyRead the docs
POST /v1/extract
curl https://api.specstream.ai/v1/extract \
  -H "Authorization: Bearer $SPECSTREAM_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"urls": ["https://en.wikipedia.org/wiki/Narbonne"], "format": "markdown"}'
Reply, cut short
{
  "data": {
    "results": [
      {
        "url": "https://en.wikipedia.org/wiki/Narbonne",
        "status": "ok",
        "title": "Narbonne - Wikipedia",
        "published": "2002-10-11T13:08:41+00:00",
        "read_from": "live",
        "text": "... ## Etymology\n\nThe etymology of the town's original name, **Narbo**, is lost in antiquity ..."
      }
    ],
    "took_ms": 441,
    "price_usd": 0.0008
  }
}

$0.80

per 1,000 pages read

Pages read, from your balance. A page with no text is not charged.

20 URLs a call
Send 1 to 20 http or https URLs. The same URL twice is read once.
PDF, Word, Excel
Documents come back as markdown or plain text, the same as a web page.

What a call returns

One result for each URL, in the order you sent them, then the request_id and took_ms of the call.

url and status
Each URL and how its read went: ok, blocked, not_found, empty, unsupported, invalid_url, or failed. ok means read and charged.
text
The page's main text. Markdown keeps headings, lists, and links. Text returns it plain.
published
The date the page says it was published, when it says one. The page's title comes with it.
read_from
live for the page now, site for the same page from the site's own API or feed, archive for its newest public web archive capture, with the capture's time in captured_at.
price_usd
The price of the pages read in this call.

How a refused page is read

A page that refuses a plain read is tried again in this order, and read_from says where the text came from.

  1. As a browser

    The request goes out again the way a browser would ask for the page.

  2. In a real browser

    If that is refused too, the page opens in a real browser and the text is read there.

  3. From the site's feed

    Next, the same page comes from the site's own API or feed, and read_from is site.

  4. From the newest public archive capture

    Last, the text comes from the newest public web archive capture. read_from is archive, and captured_at gives its time.

Extract API questions

Something we didn't answer? Talk to a human

What does a page cost?
$0.80 per 1,000 pages, from your balance, with no quote and no approval. A page that gives no text is not charged.
What if a page still can't be read?
Its status says why: blocked, not_found, empty, unsupported, invalid_url, or failed. The other URLs in the call are still returned.
Which formats does it return?
Set format to markdown, the default, which keeps headings, lists, and links, or to text for plain text. Both work for web pages, PDF, Word, and Excel.
Can I retry a call safely?
Yes. Send an Idempotency-Key header and the call is charged once per key. The same key again gets a 409.

Extract your first page

Sign up and get $10 of credit, no card.

Start freeTalk to a human