Privacy Policy

Last updated 18 July 2026.

W1 is a desktop AI coding agent that runs model inference through the hosted W1 backend (app.w1lab.com). Some data necessarily leaves your machine; this explains what, why, and how to control it.

What we collect

Model inference. Every model call is proxied through the W1 backend — that is how W1 runs without provider keys on your machine. Your prompt and the code context the agent reads or edits are sent to the backend and on to the model provider for that call.

Local run traces. Full prompts, file paths, commands, tool observations, model output, and diffs remain on your device by default.

Usage analytics. W1 sends structural product telemetry such as run status, tool names and counts, token counts, latency, cost, model route, and error categories under pseudonymous identifiers. It does not send source code, raw prompts, file paths, shell commands, model output, or tool observations to PostHog.

Website visits. If you browsed w1lab.com and then signed in from the same browser, that visit — pages seen, referrer, campaign tags — is associated with your account, so we can tell which pages and links actually bring people to W1. It is an anonymous identifier from your own browser, never your name or email. To browse without it, add ?w1_no_track=1 to any w1lab.com URL once.

What we do not collect

• Provider API keys or secrets.
• Source code or file contents in product analytics.
• Raw prompts, paths, commands, model output, or tool observations in remote traces.

Your controls

• W1_TELEMETRY_ENABLED=0 disables optional product analytics.
• Content-bearing internal observability is disabled unless explicitly enabled in an internal debugging environment.

Contact

Questions or deletion requests: support@w1lab.com.