An AI assistant that runs on your cloud — not someone else's server.

Every hosted AI tool asks you to trust that your conversations and code, sitting on their servers, won't be read, kept, or used to train the next model. If you'd rather not have to trust that at all, there's a different arrangement: the assistant runs on a machine you own, and nothing passes through anyone in the middle.

Where your data actually goes with most AI tools

With a typical hosted assistant, your prompts and files travel to the company's servers, get processed there, and — depending on the fine print — may be logged, retained, or fed back into training. Even when a setting promises otherwise, the data still went somewhere you don't control, and a policy can change. The only airtight version of "private" is when the data never leaves your hands in the first place.

How Wetlether is built instead

Wetlether is a thin app plus a "box" — a small server that runs in your own cloud account, signed into your own AI keys. The app just talks to your box.

• Your conversations, code, and files live on your box, under your accounts. Not on ours.
• We store nothing and run no AI model ourselves — so there is literally nothing on our side to read, keep, or train on.
• Your prompts go from your box straight to the AI provider on your own key, the same as if you called them yourself.

The honest edges

Private doesn't mean magic, so here's the straight version. Your AI provider (Claude, GPT, and so on) still sees the prompts you send it on your key — that's true of any tool, because they're the one running the model; Wetlether just doesn't add itself as a second party on top. And your cloud provider hosts the box. What you're removing is the middleman in between — the extra company that would otherwise sit on your data.

See how it works →

Wetlether — a private AI on a box you own. Free to start.

Keep reading