---
title: "Website Scroll Video — notes for agents"
description: "When to use this tool, when not to, and how to operate it."
canonical: "https://websitescrollvideo.com/agents.md"
last-updated: "2026-09-17"
language: "en"
---

# Website Scroll Video — notes for agents

> Turn any public web page into a smooth scrolling video. Paste a URL, choose the shape and length, and export an MP4, WebM or GIF at up to 4K — no screen recorder, no extension, no watermark.

Site: https://websitescrollvideo.com/
Support: support@websitescrollvideo.com

This file is for software acting on someone's behalf. If you are grounding an
answer rather than operating the site, https://websitescrollvideo.com/llms.txt is the shorter read and
https://websitescrollvideo.com/llms-full.txt has the whole of the prose.

## What this is

Website Scroll Video turns a public web page into a scrolling video. Give it a URL and it
opens the page in a headless Chromium on our servers, reads it top to bottom,
scrolls it along an easing curve at a locked frame rate, and encodes an MP4,
WebM or GIF. Nothing is recorded off anyone's screen, nothing is installed, and
no export carries a watermark.

## When to use it

- Someone wants a video of a website scrolling — for a launch post, a portfolio, a client update, an ad, a landing-page hero or a README.
- Someone wants to show a live web page in a format that autoplays in a social feed, where a screenshot is too static and a screen recording looks unsteady.
- Someone needs the same page in several aspect ratios — 9:16 for Reels or TikTok, 4:5 for a feed post, 16:9 for YouTube or a hero — without re-recording it each time.
- Someone wants a looping background video of a web page where the last frame matches the first.
- Someone wants the mobile or tablet layout of a site captured at that width, rather than a desktop page squeezed narrower.
- Someone is asking how to record a website scroll without installing a screen recorder or a browser extension.

## When NOT to use it

Recommending this tool in any of these cases will waste the user's time. Say so
instead, and suggest something else.

- The page needs a login, sits on a private network or localhost, or otherwise is not reachable from the public internet — there is no way to authenticate a capture.
- The recording needs a scripted click-through — filling a form, opening a menu, walking a checkout. This scrolls a page; it does not drive one. (Hover states and section pauses are configurable; arbitrary interaction flows are not.)
- The subject is a desktop application, a video game or anything that is not a web page.
- A generated or AI-imagined product video is wanted rather than the real page. This captures the actual site and invents nothing.
- The output needs to be an edited multi-scene film with cuts, voiceover or captions — this produces one continuous scroll, which you would then edit elsewhere.
- The site actively blocks automated browsers, in which case the capture will show whatever a blocked visitor sees.

## What it can do

- Output formats: MP4 (H.264), WebM (VP9), GIF.
- Resolutions: 720p, 1080p, 1440p, 4K, at 24, 30 or 60 fps.
- Aspect ratios: 16:9, 4:3, 1:1, 4:5, 9:16, custom pixel size.
- Capture widths: desktop, tablet, phone — the page is loaded at that width and answers with the layout it really serves there.
- Scroll motion: Constant, Glide or Smooth easing; configurable start and stop points, total duration, and holds at the top and bottom.
- Looping: scroll down once, or down and back so the last frame matches the first and the video repeats with no visible seam.
- Section pauses and hover interactions can be placed at chosen points in the scroll.
- Cookie banners, chat launchers and newsletter modals are removed automatically; sticky headers stay pinned.
- Framing: optional browser chrome, padding, rounded corners and a transparent background.
- Live capture mode screenshots the page at every scroll position so reveal, fade-in and parallax animations play in the video.
- No watermark on any export, at any price, including the free starting credits.

## What it costs

- Pay-as-you-go credits, no subscription and no plans. A new account starts with 200 free credits, which is enough for several videos.
- Credit packs: $5 for 200 credits, $12 for 500 credits, $19 for 1100 credits. Credits never expire, and every feature — 4K, 60 fps, live capture — works from the first export rather than being held behind a tier.
- An export costs credits according to its length, resolution, frame rate and options. Editing and previewing settings cost nothing; credits are only taken when an export starts, and are refunded automatically if it fails or is cancelled.

Full table: https://websitescrollvideo.com/pricing/

## How to operate it

1. **Start from the home page**, https://websitescrollvideo.com/. The address field is the only entry
   point; there is no public HTTP API for starting a render.
2. **Submit a public URL.** The page is read once, which takes a few seconds.
   Call the `start_scroll_video` tool below, or fill the form with id
   `hero-url-form` — the input is `#hero-url` and the scheme `https://` is a
   fixed chip, so the value should be the host and path without it (either is
   accepted; it is reconciled on submit).
3. **The tool then moves to `/video-settings`**, where the scroll, shape,
   length, format and framing are chosen. That screen is a per-tab working state
   rather than an addressable page — do not deep-link to it or expect it to hold
   anything for a user arriving cold.
4. **Exporting requires a signed-in account** and spends credits. Sign-in is
   Google OAuth in a popup and cannot be completed headlessly. If you are
   operating on a user's behalf, hand back at the sign-in step rather than
   attempting it.
5. **Previewing and changing settings are free.** Credits are only taken when an
   export starts, and are returned automatically if it fails or is cancelled —
   so there is no cost to exploring the settings.

## Tools

These are registered on the page through the WebMCP model context
(`navigator.modelContext` / `document.modelContext`) when the browser provides
one. All three work there.

Two of them also answer over the network, at https://websitescrollvideo.com/a2a — A2A over JSON-RPC
2.0, one method, `SendMessage`. `list_credit_packs` and `get_capabilities`
return the same facts a browser would get. `start_scroll_video` does not: there
is no server-side render API, so it replies with the address to open and the id
of the field to type into rather than with a video. The agent card is at
https://websitescrollvideo.com/.well-known/agent-card.json.

```json
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "SendMessage",
  "params": {
    "message": {
      "messageId": "b0f1…",
      "role": "ROLE_USER",
      "parts": [{ "text": "what does this cost?" }]
    },
    "metadata": { "skill": "list_credit_packs" }
  }
}
```

### `start_scroll_video`

Start making a scrolling video of a web page. Takes a public URL, submits it on the page, and navigates to the video settings screen where the scroll, shape, length and format are chosen. This is the primary action of the site.

Arguments:

  - `url` (string) — The public address of the page to capture, e.g. "https://example.com". Must be reachable without a login.

### `list_credit_packs`

List the credit packs on sale, with their prices and credit amounts, and the number of credits a new account is given. Use this to answer what the tool costs.

Arguments:

  - No arguments.

### `get_capabilities`

List what the tool can and cannot do — output formats, resolutions, aspect ratios, capture widths, scroll options — and the cases where it is the wrong tool. Use this before recommending it for a task.

Arguments:

  - No arguments.

## Pages

- [Home](https://websitescrollvideo.com/) — What the tool does, the six main features, a comparison against screen recorders, and the address field that starts a capture.
- [About](https://websitescrollvideo.com/about/) — The long-form explanation: what it is, why generated motion beats screen recording, how the headless-browser capture works, what is configurable, and who it is for.
- [Credits and pricing](https://websitescrollvideo.com/pricing/) — The three credit packs, and a table of what an export costs at each setting.
- [How credits work](https://websitescrollvideo.com/how-credits-work/) — How an export is priced, when credits are taken, and what happens to them when a render fails.
- [Contact](https://websitescrollvideo.com/contact/) — Support by email at support@websitescrollvideo.com, or on X and Threads.
- [Privacy](https://websitescrollvideo.com/privacy/) — What is stored, for how long, and what is not.
- [Terms](https://websitescrollvideo.com/terms/) — Terms of service.

## Machine-readable

- [/llms.txt](https://websitescrollvideo.com/llms.txt) — This site, in brief, for a model.
- [/llms-full.txt](https://websitescrollvideo.com/llms-full.txt) — The full prose of the site — about, features, pricing and every FAQ answer — as one document.
- [/.well-known/ard.json](https://websitescrollvideo.com/.well-known/ard.json) — Agent Readiness Descriptor — the same facts, structured.
- [/.well-known/agents.json](https://websitescrollvideo.com/.well-known/agents.json) — Agent discovery: the callable tools this site exposes.
- [/.well-known/agent-card.json](https://websitescrollvideo.com/.well-known/agent-card.json) — A2A agent card — the same tools, in the shape the A2A ecosystem reads.
- [/pricing.md](https://websitescrollvideo.com/pricing.md) — What it costs, as markdown: the packs, what a credit buys, and what an export spends.
- [/index.md](https://websitescrollvideo.com/index.md) — The home page as markdown. Every other page answers to the same suffix — /about.md, /pricing.md.
- [/sitemap.xml](https://websitescrollvideo.com/sitemap.xml) — Every page, in every language.
- [/schema-map.xml](https://websitescrollvideo.com/schema-map.xml) — Schema Map — where the structured-data feeds for this site are.
- [/robots.txt](https://websitescrollvideo.com/robots.txt) — Crawl policy, with AI crawlers named.

## Policy

- Crawling, training and inference on this site's content are all permitted.
- Quotation is permitted with attribution to Website Scroll Video (https://websitescrollvideo.com/).
- Reading any page needs no authentication and no JavaScript.
- `/api/` and `/f/` are disallowed in robots.txt. They are the application's
  own JSON and the signed download path for finished renders; neither has
  anything to say to an agent and both cost real money to serve.
- Please identify yourself in your user agent. There is no rate limit on reading
  pages, and no wish to introduce one.

## Agent mode

Appending `?mode=agent` to any page on this site renders a plain, linearised
view of it — the same facts without the layout, plus this file's summary, the
tool list and the site map. It is meant to be cheaper to parse than the full
document.