Skip to main content

Give AI agents a physical finger to operate any phone — robotic stylus + camera, no APIs needed.

Project description

PhysiClaw

An AI agent that physically operates a phone — the way you do.

English docs · Demo video

中文文档 · 演示视频

PhysiClaw watches a phone's screen with a camera and taps it with a stylus, working the phone the way a person would. No APIs, no OAuth, no ADB cables, nothing installed on the phone — just unlock it, set it on the desk, and let the agent work.

It's built for the everyday errands that pile up — ordering takeout, shopping for groceries, booking a ride, paying a bill, replying to a message. Anything you can do by hand on your phone, PhysiClaw can do for you.

Why a physical body?

The apps that run your daily life are closed off: most expose no public API, and simulated input — desktop automation or Android's ADB — leaves software fingerprints that anti-bot systems flag. So PhysiClaw treats the screen itself as the API: a camera reads it, a stylus performs the gestures. To the phone it's indistinguishable from a real finger — nothing to detect — and it reaches virtually any app reliably.

The trade-off is speed: a few seconds per action, in exchange for universality and reliability.

How it works

PhysiClaw has its own dedicated phone running its own WhatsApp account. Add it as a contact and chat with it like a real person — tell it what you need in plain language:

Order a latte on DoorDash.
What's the weather tomorrow?
Buy milk and eggs on Instacart.

Its runtime loops continuously, waking the agent when a scheduled task is due or the screen lights up with a new message. The agent unlocks the phone, reads your message, and does the task — pausing for confirmation when it matters, like a payment — then replies with the result, saves its memory, and exits until it's next woken.

Getting started

You'll need the PhysiClaw hardware — the arm, stylus, and camera — plus an iPhone for it to operate.

The physiclaw CLI runs on macOS, Windows, and Linux:

# Install the CLI (uv + Python 3.12 + physiclaw)
curl -fsSL https://physiclaw.ai/install.sh | bash
# Windows: iwr -useb https://physiclaw.ai/install.ps1 | iex

physiclaw models key <provider>        # add your LLM API key
physiclaw models use <provider/model>  # pick the model the agent runs on
physiclaw doctor                       # check your environment
physiclaw                              # start the server — opens the hardware-setup wizard, then runs the agent

License

MIT — both the CAD-as-code hardware and the agent software.

Project details


Release history Release notifications | RSS feed

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

physiclaw-0.3.21.tar.gz (1.9 MB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

physiclaw-0.3.21-py3-none-any.whl (589.3 kB view details)

Uploaded Python 3

File details

Details for the file physiclaw-0.3.21.tar.gz.

File metadata

  • Download URL: physiclaw-0.3.21.tar.gz
  • Upload date:
  • Size: 1.9 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.8.2

File hashes

Hashes for physiclaw-0.3.21.tar.gz
Algorithm Hash digest
SHA256 8709af82b281415b23ba8bd4a7c0f9e7807cb9019d5d87d88c1f80c255dda5c6
MD5 d7c7297572765b98e46a8eea0e9bc503
BLAKE2b-256 d6a301133d6724b8a3305b79ef23bcbb405d18ef396f589f684a6860ece52a84

See more details on using hashes here.

File details

Details for the file physiclaw-0.3.21-py3-none-any.whl.

File metadata

  • Download URL: physiclaw-0.3.21-py3-none-any.whl
  • Upload date:
  • Size: 589.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.8.2

File hashes

Hashes for physiclaw-0.3.21-py3-none-any.whl
Algorithm Hash digest
SHA256 0e80669f0922bc23c88793e94c0b416a6e5d0ed5153f03ccc119425e82f478ef
MD5 4d75daec0bb940507df5f776b5d9b96d
BLAKE2b-256 2e2ce312c4d016afbe14a3738a2a6e9fbbf2237005aa7f3f20aaa732cc7f2453

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page