Iphone control

Skill MichaelAdamGroberman/mac-mcp/skills/iphone-control

Use when the user wants to drive their iPhone via macOS iPhone Mirroring — sending text, tapping UI elements, taking screenshots of the phone, or running multi-step phone workflows. Encodes the right tool sequence (launch → focus → act) and the iPhone-Mirroring-specific quirks like coordinate translation for taps.From its SKILL.md

Install

npx -y skills add MichaelAdamGroberman/mac-mcp --skill iphone-control

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.9 KB, as published. Nobody here has run it

Driving an iPhone via mac-mcp's iphone_mirror tool

iPhone Mirroring (com.apple.ScreenContinuity, macOS Sonoma+) only works when:

Both Mac and iPhone are signed into the same Apple ID
The iPhone is unlocked or nearby
The iPhone is on the same Wi-Fi network or close enough for AirPlay

If any of those aren't true, iphone_mirror launch returns successfully but the window will show a "Cannot connect" message. There's no programmatic way to detect that — ask the user to confirm visually.

Standard workflow

For any iPhone interaction, always:

iphone_mirror action=launch — idempotent; returns already_running:true if already up.
Wait briefly (the model should not call back-to-back; insert a small await if your runtime supports it — otherwise just trust the user/runtime).
iphone_mirror action=focus — brings the window to front.
Then act.

Sending text into a focused field

iphone_mirror action=type_text text="hello world" delay_ms=12

The default 8 ms inter-keystroke delay works for iMessage and most iOS text fields. If text drops characters, raise delay_ms to 25–40.

Tapping a UI element on the phone screen

The iphone_mirror tool intentionally does NOT include tap. Use mouse_click against the mirrored window's bounds:

iphone_mirror action=get_window → returns bounds: {x, y, width, height}
Compute screen coords from the iPhone-relative tap point. iPhone Mirroring renders the phone at native resolution scaled to fit; for a tap at iPhone-coord (ix, iy) on a phone of native size (IW, IH) and window bounds (wx, wy, ww, wh):
```
sx = wx + (ix / IW) * ww
sy = wy + (iy / IH) * wh
```
mouse_click x=<sx> y=<sy>

If you only know the visual location ("tap the blue button mid-screen"), call iphone_mirror action=screenshot first, identify the location in the returned PNG, and translate from PNG coords → window coords (they're the same in this case since the screenshot is OF the window).

Capturing the iPhone screen

iphone_mirror action=screenshot

Returns a base64 PNG. If the call returns capture_failed, the user hasn't granted Screen Recording to MacMCP — invoke the mac-mcp-tcc-grant skill.

Don't

Don't try to bypass the launch step "to save a turn" — iPhone Mirroring re-attaches to the phone state on every launch and the first call without a focused window will fail silently.
Don't mouse_drag for swipe gestures on the iPhone — Sonoma's iPhone Mirroring doesn't translate them. Use mouse_scroll instead, or send the equivalent keyboard shortcut if one exists.