Security promise

In three sentences

Clawco never pays. It never messages another person and never deletes anything without your own tap on yes, and a voice command cannot give that yes. Everything it does is written to a log, with the way to undo it when there is one.

How it is enforced

Every action passes one decision step before anything runs. The kind of action comes from the app, never from the model, so what a model writes can add a refusal but cannot lift one.

How it is reached

By default Clawco listens on 127.0.0.1, the machine it runs on. You can add your own tailnet. Requests that arrive through Tailscale Funnel are refused.

Where your data lives

A Clawco you install yourself keeps your data in your user folder, on your machine.

When the agent works, the text it needs goes to the model you chose. A model on your own machine keeps that step on your machine.

This website

It sets no cookies, makes no requests to other servers and counts no visits. Fonts and scripts come from the site itself. The language toggle stores one thing, your language choice, in your own browser.

The full security text, with the tests behind each promise, is still being written.