A full build is about $112 in commodity parts, and only one of them needs to be any good: the arm. Everything else — the camera, the stylus, the mounts — is cheap, interchangeable, and forgiving. Buy the list below, or substitute freely as long as you respect the three “what matters” notes that follow.
| Part | What to get | Qty | Approx. |
|---|---|---|---|
| Arm | Paixi Kuaichaobao P25 pen plotter — X/Y gantry, Z carriage | 1 | ~$80 |
| Camera | UGREEN 1080p USB webcam, fixed focus | 1 | ~$14 |
| Stylus | Capacitive stylus, conductive fiber tip 8–10 mm | 1 | ~$1.50 |
| Camera arm | Gooseneck desk clamp, metal, ~50 cm | 1 | ~$2 |
| Phone holder | Anti-slip pad + L-shaped corner blocks | 1 set | ~$1.20 |
| USB hub | Powered USB 3.0 hub | 1 | ~$13 |
| Total (excluding your computer) | ~$112 |
The table hides the one thing worth knowing: the parts are not equally important. Here is what each one is doing, and where you can — and can’t — cut corners.
The arm — the only precision part
Everything PhysiClaw does well or badly traces back to the arm. It carries the stylus across the screen on X/Y and drops it on Z. A tap that lands a few millimeters off is the arm, not the agent. The P25 is cheap, rigid enough for a phone-sized work area, and speaks standard G-code — but any GRBL- or FluidNC-compatible X/Y plotter works. Spend your attention here.
The Z axis is a solenoid
The stock P25 ships with a small servo to lift its pen. PhysiClaw replaces that with a solenoid — an electromagnet that snaps the stylus down to tap and releases to lift. A solenoid gives a crisp, repeatable strike a slow servo can’t, which is why a long-press reads differently from a tap. The firmware drives it over the board’s spindle output, not a servo channel.
The stylus must be soft conductive fiber
To the phone the stylus is a fingertip, so it has to fool a capacitive touchscreen. Use a soft conductive-fiber tip, 8–10 mm across. Hard rubber or plastic tips register unreliably and can scratch the glass. This is a $1.50 part; don’t overthink it, but don’t swap in a “precision” hard nib.
One camera, overhead
A single fixed-focus USB webcam looks straight down at the screen — there is no second camera. It serves two jobs: a clean overhead read of the screen with the stylus parked away, and a check of where the stylus is sitting before a tap. 1080p is plenty; fixed focus is preferred so the image doesn’t hunt.
The rest of the list just holds the three important parts in place:
