> ## Documentation Index
> Fetch the complete documentation index at: https://docs.browserwire.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Recording

> How BrowserWire captures browser sessions as structured training data.

A **recording** is BrowserWire's raw training data — a captured browser session that the AI training pipeline uses to learn a site's structure. You create recordings with the `bw record` command.

## How recording works

When you run `bw record <url>`, BrowserWire:

1. **Launches a Chrome browser** via Playwright and navigates to the URL you specified.
2. **Waits for you to start** — you can browse freely before recording begins. Press **Enter** to start.
3. **Captures DOM events** — once recording starts, BrowserWire injects [rrweb](https://www.rrweb.io/) into the page. Every DOM mutation, scroll, click, navigation, and form interaction is captured as a structured event stream in real time.
4. **Follows navigations** — if you click a link that opens a new tab, BrowserWire redirects it into the same page so the recording stays continuous.
5. **Stops on Enter** — press **Enter** again to stop recording. BrowserWire reloads the page at the start of recording to get a clean initial snapshot.

## What gets captured

| Data                  | Description                                                        |
| --------------------- | ------------------------------------------------------------------ |
| **DOM snapshots**     | Full page structure at recording start, plus incremental mutations |
| **User interactions** | Clicks, keypresses, form fills, scrolls                            |
| **Navigations**       | URL changes, page loads, SPA route transitions                     |
| **Event metadata**    | Timestamps, event types, element targets                           |

Recordings are stored as JSON arrays of rrweb events. They do **not** contain screenshots, network requests, or cookies.

## Upload and training

After you stop recording, the CLI:

1. Authenticates (prompts for login if your token has expired)
2. Uploads the recording events to BrowserWire's cloud storage
3. Triggers a training run automatically
4. Opens the dashboard so you can monitor progress

The recording metadata (domain, event count, size) is stored in your BrowserWire account and visible in the [dashboard](/guides/dashboard).

## Tips for better recordings

* **Cover key paths** — focus on the flows you want to automate (login, search, checkout, etc.)
* **Navigate slowly** — let pages fully load before interacting. The AI needs stable DOM snapshots.
* **Interact with dynamic UI** — open dropdowns, expand accordions, trigger modals. State changes help the AI discover more actions.
* **Use a real account** — if the site requires authentication, log in during the recording so the AI can see authenticated pages.
* **Re-record to improve** — you can record the same site multiple times. Each recording triggers a new training run that can improve or update the manifest.

<Note>
  Recordings are uploaded to BrowserWire's cloud for processing. They contain DOM structure and interaction events from the sites you record. See [Privacy](/reference/privacy) for details.
</Note>