An agent that can see and touch
The AI that interacts with you in the real world.
Give your agent a phone. It handles the rest.
// How it works
PhysiClaw gives your agent a body — a hand and an eye. A gantry moves across the plane, a stylus taps the screen, and a camera watches what happens.
Message PhysiClaw from any chat app, like texting a friend. With its own phone, it reads and replies on its own — no OAuth, no API to wire up.
It works the phone with a stylus — the same taps, swipes, and scrolls you'd use. It opens the right app for the task and navigates by hand, no special access required.
A camera is its eye. After every action it checks what changed — did it land, did the job get done — and glances back now and then so nothing slips by.
It never runs commands on your computer or touches your files, and keeps only the memory and cron jobs it needs. Its whole world is the phone in front of it.
// Why PhysiClaw
The services you live in weren't built for your agent. WhatsApp, WeChat, Amazon, Taobao, Uber, Didi, Airbnb — their APIs are gated. An agent that can only call APIs is locked out of the apps that matter most.
A body changes that. Tapping the phone like you do, it gets in everywhere. To the phone, it's a finger on screen. The data wall can only be broken physically.
Both the hardware and the agent are open source. Buy the parts, assemble your own, and let your personal assistant work for you.