Skip to main content

​
What gets uploaded

When you run bw record, the captured DOM events (rrweb recordings) are uploaded to BrowserWire’s cloud storage for training. This includes:
  • DOM structure and mutations
  • User interactions (clicks, keystrokes, form fills)
  • Navigation events and URL changes
Recordings are stored in your account and used solely to generate manifests. They are not shared with other users or third parties.

​
What stays local

  • Session tokens — stored at ~/.config/browserwire/config.json with 0600 permissions
  • API execution — when you run bw run, all API calls are served locally through a browser on your machine. No execution data is sent to the cloud.
  • Browser instances — Playwright browsers launched by bw record and bw run run entirely on your machine

​
What’s sent to the cloud

DataWhenPurpose
rrweb recording eventsbw record (after you stop)Training input for manifest generation
Recording metadata (domain, event count, size)bw recordDashboard display and training orchestration
Execution metrics (route, duration, success/failure)bw run (fire-and-forget)Usage analytics for your dashboard
Authentication tokensbw loginAccount authentication via Clerk

​
LLM processing

Your recordings are processed by a vision-capable LLM as part of the training pipeline. The LLM runs server-side in BrowserWire’s infrastructure. The recording events are sent to the LLM provider (e.g., OpenAI, Anthropic) as part of the training process.

​
DOM snapshot recording

BrowserWire uses rrweb (MIT license) to capture DOM snapshots during recording. The library runs client-side in the Playwright browser and streams events to the CLI process.

​
No background collection

BrowserWire does not run in the background. Recording only happens when you explicitly run bw record and press Enter to start. The CLI exits after each command completes.
If you have questions about privacy or data handling, open an issue at github.com/gearsec/browserwire.