Overview
Drive a desktop environment's screen, mouse and keyboard by hand
fuse computer drives a desktop environment interactively: capture the
screen, click, type, send key combos. It exists so a desktop session can be
inspected by a human without writing an agent loop — the commands call the
same POST /v1/environments/{id}/computer endpoint an agent loop uses through
the SDKs.
It requires an environment booted from a desktop image (see desktop environments); on any other image the server answers 503 with a reason.
Commands
| Command | Purpose |
|---|---|
fuse computer screenshot |
Capture the screen to a PNG |
fuse computer click |
Left-click at a coordinate |
fuse computer type |
Type text into the focused window |
fuse computer key |
Send a key combo |
Unlike fuse environment exec, these
commands accept API keys: an action here is a mouse click on a display the
image chose to run, not a root shell.
The live desktop also appears on
fuse environment get as a desktop row
with its geometry and up/down state, next to the endpoint rows.