Screenshots
screenshot (and jump screenshot) capture what is on the node’s screen, so the agent must
run in the logged-in user’s desktop session, not as a service or over SSH.
| OS | Notes |
|---|---|
| Windows | Works while the session is unlocked and connected. A locked screen or a disconnected Remote Desktop session can’t be captured. |
| macOS | The first screenshot asks the user on the node to allow Screen Recording for jump-agent (System Settings > Privacy & Security > Screen & System Audio Recording). Restart jump-agent after allowing it. Unsigned builds must be allowed again after each update. |
| Linux | X11 sessions only. Wayland (the default on recent Ubuntu and Fedora) isn’t supported yet: choose an X11 session such as “Ubuntu on Xorg” at the login screen. The agent needs DISPLAY set, which it is when started from the desktop session. |
Displays are numbered from 0, the main display. Scaled-down screenshots report the display’s real size, so your AI can ask for a region, in display pixels, to read small text.