GhostDesk
未认领Give any AI agent a full desktop — it sees the screen, clicks, types, and runs apps like a human. Automate anything with a UI: browsers, legacy software, internal tools. No API needed. One Docker command.
工具（14 个）
app_launch
Start a GUI application by name
app_list
List the GUI applications installed on the desktop
app_running
List the application windows currently open — call before applaunch to avoid relaunching an app that is already there
app_status
Check if an application is running and read its logs
clipboard_get
Read clipboard contents
clipboard_set
Write to clipboard
key_press
Press keys or combos (ctrl+c, alt+F4, Return...)
key_type
Type text with realistic per-character delays
mouse_click
Click at coordinates
mouse_double_click
Double-click at coordinates
mouse_drag
Drag from one position to another
mouse_move
Move the cursor to coordinates without clicking — reveals hover-only menus, tooltips, and CSS :hover states (e.g. Gmail action bar)
mouse_scroll
Scroll in any direction (up/down/left/right)
screen_shot
Capture the screen as a WebP image (pass format="png" for lossless). Pass region= to crop to a sub-rectangle at native resolution. Tune quality= (1-100, default 50 — invisible on UI content, ~50% smaller than 80; raise for fine fonts or design surfaces). Set stabilize=False to skip page stabilizatio