What problem does it solve?
It lets an agent operate the user's real local computer—taking screenshots, clicking, typing, scrolling, and inspecting native app UI—through a connected A0 CLI or Launcher host gateway, without confusing the host machine with the internal Agent Zero Docker desktop.
Core Features & Use Cases
- Remote desktop control: Start sessions, capture screenshots, move, click, scroll, send keys, and type text on the host machine via the
computer_use_remote tool.
- Structured UI targeting: Use
list_windows, get_window_state, and element_action with background dispatch when the backend advertises accessibility-tree support (AT-SPI on Linux, Accessibility on macOS, UIA on Windows).
- Clear boundaries: Distinguishes host control from the internal Linux desktop and from browser automation, directing browser tasks to the browser tool instead.
- Use Case: Ask the agent to open a native desktop application on your Ubuntu Wayland machine, navigate its menus, and fill in a dialog, with each step verified against fresh screenshots.
Quick Start
Ask the agent to take a screenshot of your local computer screen and click a specific button in a native application window.