Your apps, used in the background
Coding agents are good at code and blind to everything else: the simulator that shows the bug, the design tool with the spec, the settings screen a test depends on. ace gives every agent the same computer-use tools, whichever CLI runs it, so Claude Code, Codex, Cursor and the rest can open an app, read what is on screen and act on it.
Agents read each app through the system's accessibility interface, so they see real buttons, fields and labels instead of guessing from pixels. They work in the background: your mouse and keyboard stay yours while an agent fills a form in another app.
You stay in control
Every app an agent is using shows in ace with the agent's name beside it, a live view, and Take over and Stop. Take over hands the app back to you mid-task; Stop ends the agent's session with it at once.
Computer use is off until you turn it on, with one switch. You decide which apps agents can never use, such as mail or a password manager, and ace refuses those apps whatever the agent asks.
Permissions stay with your CLI
ace never adds its own approval layer. Each agent's CLI decides what runs, in the permission mode you already chose for it, and asks you the way it always does. When an agent waits on an approval, its thread says Approve, on your desktop and on your phone.