For AI assistants

A download tool an assistant can call

GrabGrab is a normal website and a simple HTTP API. An assistant does not have to drive a browser. It sends an X link, or a TikTok, Vimeo, or Facebook link, reads the JSON, and saves the file.

  1. Step 1

    Copy a public video URL from X, Twitter, TikTok, Vimeo, or Facebook.

  2. Step 2

    POST it to https://www.grabgrab.fun/api/download as JSON with the fields url and videoQuality. 1080 is a good default.

  3. Step 3

    If success is true and type is direct, download the url field with curl -L. If the CDN blocks that, use the signed proxyUrl on grabgrab.fun instead.

  4. Step 4

    Tell the person the filename. If the response is an error, show it. Do not retry YouTube — it is refused on purpose.

The public API has a per-IP limit. When the human check is enabled, browser visitors complete it and assistants keep using this direct call with the smaller limit. Do not invent a secret key. Full notes are in /llms.txt.

Read llms.txt

Using GrabGrab from an assistant

Does the assistant need an account?

No. The public download API does not require a GrabGrab account.

Will a YouTube URL work if the user insists?

No. The API returns youtube_not_supported and does not try the download.

What if the request is rate limited?

The API returns HTTP 429 and code rate_limit. Wait a few minutes. Do not hammer it.

Where is the machine-readable description?

https://www.grabgrab.fun/llms.txt and https://www.grabgrab.fun/llms-full.txt