# Auto Capture by Tella: AI screen recording agent

> Auto Capture is Tella's agentic screen recorder, a camera for browser agents. Give an AI agent a URL and it opens the page in its own browser, explores it, plans a walkthrough, checks every step works, and films it with an animated cursor, with no one at the keyboard. The result is a real Tella recording of the live product that an editing agent can then cut, zoom, caption and share. Auto Capture is in beta on paid plans.

Page: https://www.tella.com/auto-capture

## How a take works

Each request is a take that films one page or flow, and usually runs for several minutes.

1. **Explores.** Opens the URL in a fresh browser and decides what to show and in what order. With a login saved in Tella, it signs in here, off camera.
2. **Scripts and checks.** Writes the walkthrough as a script and runs it to make sure every step works. Anything a check saves on the account is put back first, so the recording starts from the account as it was; before a change it can't undo, it stops and asks.
3. **Records.** Plays the script in a clean browser and films it in 4K (a 1920×1080 page at 2× density) at 60 fps, moving an animated cursor to each thing it points at.
4. **Delivers.** Hands back the recording with a handoff note (what it shows, what happened off camera, where it departed from the request), chapter markers, and suggested cuts for page loads and long pauses.

## Who can start a take

- **The Tella Agent** in Tella's editor. It places the take in the video you're editing, for example as b-roll over the part where you talk about that page. It can also start a take on its own when an edit needs footage your video doesn't have.
- **Any AI app connected to Tella's MCP server** (`https://api.tella.com/mcp`): Claude, ChatGPT, Claude Code, Codex, Cursor and others. People sign in with their Tella account; scripts and CI jobs send a Tella API key (Settings → API Keys) as the bearer token. Tools: `create_auto_capture` (with `targetUrl` and `instructions`; returns a `sourceId`), `get_auto_capture` (poll every 30–60 seconds until `delivered`, `needs_guidance`, `failed` or `cancelled`), `revise_auto_capture` (change a delivered take, or answer its question, without starting over), `cancel_auto_capture`. Place a delivered take with `create_video` or `upload_clip`.

## If you're an AI agent

- Pass on what your human wants (who the video is for, what to show, skip or end on) as `instructions`, plus what you know about the account: whether it's a sandbox, which record to use, how a change is undone, what must not change. Don't write a click script: the capture agent explores the page and picks the steps itself.
- If the demo saves something and you don't know whether the account may change, ask your human first. When `get_auto_capture` returns `accountChanges`, tell them what the take left changed.
- Never ask for a password or sign-in code in the chat. Logins are saved in Tella under **Settings → Credentials**, typed in or picked from 1Password (which also fills in one-time codes); other codes are entered on a Tella page.
- Step-by-step guide: [Record a product demo with an AI agent](https://www.tella.com/edit-with-ai/record-a-product-demo-with-ai.md).

## What it can and can't record

- Records public `http` and `https` pages, and signed-in web apps on a public domain with a saved login. Password fields are always masked.
- Can't record `localhost`, IP addresses, intranet hosts, URLs with a username or password in them, desktop apps, mobile apps, or sites that block automated browsers with a "verify you're human" check.
- Doesn't add AI narration: Tella doesn't generate voiceovers. Add a recorded voiceover as a sound, or narrate a recording yourself and use the take as b-roll.

## Use cases

- [Pricing page walkthrough](https://www.tella.com/auto-capture/pricing-page-walkthrough-video.md): a short tour of your plans, filmed from a URL.
- [Feature launch videos](https://www.tella.com/auto-capture/feature-launch-video.md): film a new feature without booking anyone's time.
- [Web page b-roll](https://www.tella.com/auto-capture/website-b-roll.md): footage of the page you talk about in a recording.
- [Help center videos](https://www.tella.com/auto-capture/make-help-center-videos.md): a video for each help article, recorded from its written steps.
- [Customer onboarding video](https://www.tella.com/auto-capture/customer-onboarding-video.md): a new customer's first steps in your app.
- [Demo for every release](https://www.tella.com/auto-capture/demo-video-for-every-release.md): one take per new page or flow, started by a person or by an agent in a release workflow.
- [Coding agent demos](https://www.tella.com/auto-capture/record-web-app-demo-from-claude-code.md): ship a feature, then have your coding agent film it.
- [Demos behind a login](https://www.tella.com/auto-capture/record-a-demo-behind-a-login.md): your signed-in product, two-step codes included.
- [Multi-page product tour](https://www.tella.com/auto-capture/multi-page-product-tour.md): one take per feature, joined into one tour video.

- A demo on every pull request: a coding agent in CI films the change on its preview deploy (it must be on a public domain, not localhost) and posts the video to the PR.
- Whole tutorials from one prompt: ask the Tella Agent for a tutorial on a new feature; Auto Capture records the walkthrough and the agent edits it with zooms, captions and chapters.
Every Auto Capture use case is listed in [AI features](https://www.tella.com/ai.md).

## Get started

1. Check access: Auto Capture is in beta and included with paid plans; a free trial can't start a take.
2. For pages behind a sign-in, save a demo login under **Settings → Credentials**. Public pages need nothing.
3. Ask the Tella Agent or a connected AI app for a take, with the page's full URL, who the demo is for and where it should end.
4. Read the handoff note, then edit and share the video like any Tella recording.

Docs: [Get started with Auto Capture](https://www.tella.com/help/integrations/auto-capture-get-started), [Example prompts and videos](https://www.tella.com/help/integrations/auto-capture-example-prompts), [How Auto Capture works](https://www.tella.com/help/integrations/auto-capture).

## How it differs from other tools

- **AI video generators** (Sora, Veo) invent footage. Auto Capture films your real product.
- **Avatar and AI-narrated video makers** put a presenter or voiceover over slides or screenshots. Auto Capture records a real walkthrough.
- **Interactive click-through demo tools** capture screens you click through yourself and produce a clickable demo, not a video.
- **Playwright or Puppeteer recordings** film a script you write and maintain. Auto Capture writes and tests the script from a plain-language request and delivers the take into Tella's editor with chapters and suggested cuts.

## Pricing

In beta, included with paid plans. Each take uses the workspace's credits (the same balance the Tella Agent spends), charged when the take ends, including takes that end with a question, fail or are cancelled.

## Related

- [Auto Capture vs Arcade, Supademo, Guidde and Playwright](https://www.tella.com/alternatives/tella-auto-capture-vs-arcade-supademo-guidde.md)
- [What is agentic recording?](https://www.tella.com/edit-with-ai/agentic-recording.md)
- [Tella Agent: agentic editing](https://www.tella.com/agent/llms.txt)
- [How Auto Capture works (help)](https://www.tella.com/help/integrations/auto-capture)
- [Let Auto Capture past your firewall or bot protection](https://www.tella.com/help/integrations/auto-capture-site-access)
- [Let Auto Capture sign in to your product](https://www.tella.com/help/integrations/auto-capture-credentials)
- [Auto Capture security](https://www.tella.com/legal/auto-capture-security)
- [Tella MCP server](https://www.tella.com/docs/mcp-server)