Browser and desktop clients

Device agent and controller

The device agent runs on the computer being operated. It brings the device online and provides supported capabilities. The controller runs on the computer in front of you and lets you select devices, see screens and send actions.

Whether you use a browser, desktop client or AI, first install the agent and bring the target online. Changing clients does not expand your access permissions.

Work in your browser

Open Workspace devices, select an online device, then choose the desktop, files or terminal tool. Your browser uses your signed-in identity and workspace permissions. You do not need an MCP key.

This is the easiest place to make a first connection. Downloaded files go to the computer running your browser; terminal commands run on the remote device.

Use the unified client

In Downloads, choose “Install with an account” and your system. Download a published unified client publicly, then sign in inside the app and choose a workspace to access authorized online devices. No controller key is required. Use the browser if no release is available for your system.

On Windows/macOS, one client accesses other devices and offers a setting to enable remote access to this device. Signing in registers the client without automatically enabling remote access; enabling it requires group selection and system authorization. Linux provides a CLI, and iOS/iPadOS only support outgoing connections. See Downloads for platform capabilities.

Connect a target without signing in on it

Use No-account installation in the workspace. An admin obtains a group-specific installer, command, or QR code, then installs and configures the target without an account sign-in. This preserves the existing Agent deployment flow for servers and managed devices. Remote operators still sign in and need device permissions.

Work with AI

Built-in Chat uses your workspace identity and its configured Chat permissions. External AI clients connect through MCP with a separately configured access key.

You can operate devices yourself or ask AI to handle part of a task. Before returning to manual operation, confirm the AI task has paused or stopped, then open the desktop to avoid simultaneous input.