Request parameters

// Ads blocked - block_ads defaults to true
{ "url": "https://example.com" }

// Ads NOT blocked - the field was sent as false
{ "url": "https://example.com", "block_ads": false }

// Ads blocked - null is treated as "not sent"
{ "url": "https://example.com", "block_ads": null }

Field names are snake_case. Unknown fields are ignored rather than rejected, so a request written against a newer version of this page still works — but a misspelled field is silently ignored too, which is worth remembering when a setting appears to have no effect.

An omitted format resolves from the category rather than to one fixed value: image gives png, video gives mp4, pdf gives image_pdf, content gives markdown.

Allowed formats for an image capture:

For a video capture:

A video records the page scrolling from the top and stops at the bottom of the page or after 30 seconds, whichever comes first. A lower scrolling_delay_ms or a smaller zoom fits more of a long page in.

For a PDF capture:

For a page-content capture (see Page content):

Set media_category to content (or just send "format": "markdown") to get the page's readable main content instead of a picture: the article, docs or product text without navigation, footers, ads, cookie banners or scripts. markdown keeps headings, lists, links and tables; text is the same content without markup.

{
  "url": "https://example.com/blog/post",
  "media_category": "content",
  "format": "markdown",
  "response_type": "by_format"
}

The page is loaded in a real browser, exactly like a screenshot, so pages built with JavaScript work, and proxy, headers and cookies apply. A content capture counts as one capture against the same allowance as an image and is charged only when it succeeds.

With response_type: by_format the answer is the text itself (text/markdown; charset=utf-8 or text/plain; charset=utf-8); with json you get a download_url as usual.

Send a page_source block to save the page's own source alongside the capture — one run, two files. Sending the block is what turns it on; there is nothing else to set.

The source comes back as its own page_source block on the response, with adownload_url of its own that expires when the capture's does. Under response_type: cloud_storage it is pushed to your bucket as a second object beside the capture.

What to capture, and what to do to the page before the shot. Every field is optional and read by presence: an empty or whitespace value means “no selector”, never “an empty selector”. Each may be at most 512 characters, except hide, which takes up to 25 selectors in 2000 characters.

Omitting the block leaves the browser’s own behaviour untouched — it does not report a position of 0,0. This sets what the page is told when it asks the browser where it is; it does not change where the request comes from. For that, use proxy.

Only the location is named. Which proxy is used and what its credentials are belong to the platform and never travel over this API.

A capture carries no timestamp strip unless the request sends this block. Sending the block at all is what turns one on, so "timestamp_watermark": {} is already a complete instruction.

// A complete instruction: the block being present turns the strip on
{ "timestamp_watermark": {} }

// Same strip, moved and recoloured
{
  "timestamp_watermark": {
    "location": "bottom",
    "font_color": "#FFFFFF",
    "background_color": "#111111"
  }
}

// Colours kept, strip switched off
{ "timestamp_watermark": { "enabled": false } }

Required for response_type: cloud_storage and rejected for the other two modes. The fields, the credential handling and the failure modes are covered in full on Deliver to your S3 bucket.

A few fields are accepted by the schema but refused by your plan. They answer 403 feature_not_available with the field named, rather than being silently dropped:

• headers and cookies — every paid plan, not the free trial
• proxy — every paid plan, not the free trial
• callback_url — every paid plan, not the free trial

Which plan includes which is set out on the pricing page. A refusal here is a plan question, not a request bug — the same body will work unchanged after an upgrade.