I let an AI agent drive my remote desktop. It can't touch my passwords.

A developer built DeskVNCViewer, a native VNC/RDP client that lets an AI agent drive a remote desktop session while keeping passwords and credentials out of the agent's reach. The agent can view, click, type and navigate, but stops at admin login prompts and waits for the human to enter credentials; control is granted through a time-limited lease that can be reclaimed instantly, and connections to someone else's PC require a short code and explicit approval. The design keeps the human in control by default, with the agent receiving only what is explicitly granted for a limited period.

It's 11 PM. Three servers need the same patch, and I'm alt-tabbing between four remote windows. Two are RDP, two are VNC. So I tried something different: I handed one session to an AI agent and told it to run through an installer wizard. It clicked through every screen without me touching anything. Then it hit the admin login prompt and stopped. It didn't try to type a password. It waited for me. I typed it myself, and the agent carried on. That's the design idea behind DeskVNCViewer, a native VNC/RDP client. How the agent boundary works No secrets for the agent. It can view, click, type, and navigate, but passwords and credentials stay with you. Time-limited lease. You grant control for a set period, and when it ends, control returns to you. Instant take-back. You can reclaim the session at any moment. Approval-based help. For someone else's PC, they generate a short code and must approve the connection before you see anything. By default you're in control. The agent only gets what you explicitly grant, for a limited time. Everyday features