To give a command, users hold a key and speak. The agent reads the screen, plans the steps, then clicks and types in the user's real browser and apps. In the browser it works through the page's DOM using an extension. On desktop apps it combines the OS accessibility tree with screenshots. It asks for confirmation before anything is sent or deleted, and it only listens while the key is held down. The company is based in Jönköping. Philip Alm and David Bacelj founded it in early 2024, and it has raised SEK 19 million (about $2 million) in seed funding. That comes on top of $700,000 raised in 2024 under the name VISS.AI. Starbright Invest and Spintop Ventures both put money in again. A public release is planned for later in 2026. Because the agent works inside sessions where the user is already logged in, it can reach supplier portals and internal tools that have no API. The cost is that reliability now depends on how well it reads the screen and on human approval steps. For founders, the open question is whether workflows learned for each user keep customers around before the model vendors ship computer use as a standard feature.