Tools

The MCP server provides 15 tools. Coordinates are in points, matching the pixels of a full-screen screenshot (430×932 on an iPhone 15 Pro Max), so the agent can tap what it reads off a screenshot.

Looking at the screen

Tool	What it does
iphone_status	WDA ready or not, locked or not, foreground app, screen size
iphone_screenshot	Screenshot; with x, y, width, height it returns an enlarged region for small text and icons
iphone_list_elements	Elements with text: type, text, center, size; filter by text or type
iphone_wait_for	Wait until some text appears (page loaded, dialog shown); text inside input fields doesn't count

Acting

Tool	What it does
iphone_tap	Tap coordinates
iphone_tap_element	Tap by text; if the center is covered by a floating bar or the keyboard it taps a visible part, and if it's fully covered it doesn't tap
iphone_long_press	Long press
iphone_swipe	Swipe; returns once the list has stopped scrolling
iphone_type_text	Type into the focused field (Chinese supported), read it back and fix dropped characters; submit presses Return
iphone_press_button	home, volume_up, volume_down, lock
iphone_launch_app / iphone_terminate_app	Launch / quit an app by name (e.g. 设置) or bundle id
iphone_open_url	Open a link or URL scheme
iphone_list_apps	Installed apps
iphone_batch	Run up to 20 steps in one call, stopping at the first error

See the screen after an action: then

Every action tool accepts then: it waits for the screen to settle and returns it, saving a round trip.

Parameter	Meaning
then	elements (element list) or screenshot
then_filter	Only elements containing this text (or of this type)
then_region	Only an enlarged [x, y, width, height] region of the screenshot, e.g. a game board
then_settle	How long to wait for the screen to settle, in seconds (default 1.5, max 10)

The result says how long it waited, whether the page is probably still loading, and whether a dialog is in the way. Even without then, it warns when a system alert covers the screen.

All parameters: docs/tools.md in the repository.