What you'll build

Before you buy anything, here’s the whole picture: what the finished rig looks like, what it costs, and the handful of steps between a box of parts and an agent tapping your phone.

            ┌─ overhead camera (on a gooseneck, ~25 cm up)
            │        looks straight down at the screen
            ▼
        ┌────────┐
        │ camera │
        └────────┘
            ╎  sees
            ▼
   ┌───────────────────────┐        ┌──────────────┐
   │  pen-plotter arm      │◄──USB──│  your Mac    │
   │  X/Y gantry           │  12 V  │  physiclaw   │
   │  Z = solenoid + stylus│◄──pwr──│  server      │
   └─────────┬─────────────┘        └──────────────┘
             │ taps
             ▼
      ┌──────────────┐
      │  the phone   │  seated face-up in a corner holder
      │  (unlocked)  │
      └──────────────┘

What to expect

≈ $112

Off-the-shelf parts. The arm is a consumer pen plotter; the camera is an ordinary USB webcam.

≈ an afternoon

Assembly is mechanical — no soldering, no custom boards. Calibration is a guided wizard.

No special skills

If you can follow a wiring diagram and clamp a gooseneck, you can build one.

macOS today

macOS is the supported platform; Windows works for most of the flow. (Serial ports differ.)

The road ahead

The rest of the docs are this build, in order. Each step hands off cleanly to the next:

Build the hardware — order the parts, then assemble the arm, stylus, camera, and phone holder.
Install the software — one command installs the physiclaw CLI; physiclaw doctor checks your machine.
Prepare the phone — turn on AssistiveTouch and add the three iOS Shortcuts. Five minutes, once.
Calibrate — a guided wizard teaches the arm and camera exactly where the screen is, so taps land where the agent aims.
Run your first task — connect your agent and give it a goal in plain language. Watch it look, decide, and tap.