Agent-readable docs index: /llms.txt. Full docs in one file: /llms-full.txt. Download /docs.zip to grep all markdown files locally.

swipe

swipeTOOL
Swipe a finger like a person from (x1,y1) to (x2,y2) in screenshot coordinates: fast, lifted while moving. For TikTok, feeds, lists, carousels, Home Screen pages
The start point and the length (the distance to the end point) are yours; speed, easing and a slight sideways bow are random on every call, so no two swipes are alike. Vary the start and the length between calls: the same swipe every time is not human.
  • Fast start, and the finger lifts while still moving, so lists keep scrolling (momentum). Content moves further than the swipe length. For an exact distance use scroll, or --duration 400 and more, which is slower but still lifts moving.
  • TikTok: swipe 200 700 200 250 swipes up from the lower middle: the next video. Both points must be on screen.
  • Home Screen pages: start in empty space or on an app icon, not on a widget and not on the AssistiveTouch button.
  • XCTest mode is fastest. Bluetooth mode needs ~100 ms more for the pointer to reach the start point; its first swipe after a connection measures the pointer speed (~1 s, one log line, the screen should be still).
  • Items that need a long press first (Home Screen icons) use hold-drag; so does exact drag and drop. Do not use it to go Home: ghosthand key home.
CLI flags in this text are arguments: --duration = durationMs, --screenshot = screenshot, --screentext = screentext, --language = language, --scale = scale, --wait-text = waitText, --timeout-ms = timeoutMs
Examples (arguments): {"lease":"studio:42","x1":200,"x2":200,"y1":700,"y2":250} {"durationMs":150,"lease":"studio:42","x1":340,"x2":60,"y1":500,"y2":500}
Parameters
durationMs?number
Fixed duration (default: random 80-180, grows with the distance)
language?string
OCR language, e.g. it-IT; en-US is always included
lease *string
Lease id exactly as acquire printed it, e.g. "studio:42"
mac?string
Mac slug from list, e.g. "studio". Needed when several Macs are online
scale?number
Image and coordinate size as a fraction of native pixels (default: screen points; 1 = native)
screenshot?boolean
Return the screen image after the input, once the screen is still
screentext?boolean
Return the OCR text of the screen after the input (x,y wxh text lines)
timeoutMs?number
With waitText: deadline in ms, 1...60000 (default: 5000)
waitText?string
After the input, poll OCR until a line contains this (case-insensitive); never repeat input
x1 *number
Start x
x2 *number
End x
y1 *number
Start y
y2 *number
End y
Request example
{
  "method": "tools/call",
  "params": {
    "name": "swipe",
    "arguments": {
      "durationMs": 0,
      "language": "string",
      "lease": "string",
      "mac": "string",
      "scale": 0,
      "screenshot": true,
      "screentext": true,
      "timeoutMs": 0,
      "waitText": "string",
      "x1": 0,
      "x2": 0,
      "y1": 0,
      "y2": 0
    }
  }
}