Your keys, your data: how Vynix runs AI
Bring your own AI keys, keep your secrets encrypted at rest, and decide exactly what context leaves your page. A look at the privacy model.
AI features are only useful if you trust where the data goes. Vynix is built so the sensitive parts stay yours: your model keys, your captured context, your choice of what is shared.
Bring your own keys
Connect your own provider key and AI calls run on your account, under your limits and your billing. Keys are encrypted at rest, never returned to the browser, and never logged in plain text.
Capture with intent
- The widget records the element, page, and the console and network evidence you choose to include.
- Password fields and values you did not deliberately capture are left out.
- Screenshots are encrypted before they are stored.
Encrypted at rest, never in the browser
Provider keys are hashed or encrypted at rest and are never returned to the client, so a key you paste in once does not leak back out through the dashboard or the network tab. AI calls run server-side on your behalf, and every run is metered so you can see exactly what it cost.
Diagnostics are metadata, not payloads
Runtime capture is intentionally shallow. Vynix records that a request failed and its status, not the request or response body, and it strips every captured URL down to its path so query strings never leave the page. The agent gets enough to reason about a failure without you shipping tokens, headers, or personal data along with it.
Off with one attribute
If even metadata is more than you want to collect on a given page, diagnostics can be disabled with a single data attribute on the widget. The annotation still captures the element and intent; the runtime evidence simply is not gathered. You set the ceiling, per site.
Why this matters for teams
A feedback tool sits on your live product, often in front of real users. That is exactly where a careless capture would hurt most. Building the privacy model first, your keys, metadata-only diagnostics, encryption at rest, is what makes it safe to leave the widget running everywhere it is useful.
The result is context rich enough for an agent to act on, without turning your feedback tool into a data-exposure risk. Useful and careful are not a trade-off here.