The short version: your screen, your keystrokes and what you ask the assistant stay between your own devices. We cannot read them, because they are encrypted end to end and the key never leaves your devices.
Screen video, pointer and keyboard input, clipboard text, your task history, the assistant's memory and your schedules are stored on the computer you control and sent only to devices you have paired. They are encrypted with keys generated on your devices during pairing.
When you are away from your own network, traffic goes through our relay so the two devices can find each other. The relay forwards encrypted bytes. It sees the size and timing of those bytes, the IP addresses of the two ends, and a random routing token that lets it match one connection to the other. It cannot decrypt anything, and the code that forwards is structurally unable to reach the code that does crypto — a test fails the build if that ever changes.
The assistant runs on your computer and uses the model provider you configure with your own key — Anthropic, OpenAI, or a local one. Your prompts and the screenshots it needs go to that provider under your own account and their terms. We do not receive them, and we do not resell model usage.
Every paired device is listed on the computer and can be revoked there, which immediately cuts any live session. Deleting the helper removes the local data it created. The assistant cannot write files unless you give a task that reach, and what it is allowed to do is bounded before it starts.
Questions or a deletion request: support. This page changes when the product does; the date below is the last change.
Last updated 20 September 2026 · Innoedge Technologies Limited