Opposable launches tool letting AI agents operate Android and iPhone screens
Overview
Opposable launches a tool that lets AI agents tap, swipe, and type on connected Android phones, and control iPhones through a companion computer.
The Android version uses accessibility services to read the screen and sends screenshots to a vision-language model, GUI-Owl 1.5 2B, run locally via llama.cpp. Installing it requires sideloading an APK and manually granting accessibility access. Opposable's website warns that "whoever holds the token can operate the phone, so treat it like a password."
The Register tested Opposable's website examples. A dark-mode change and a five-minute timer took 50 seconds and 17 tool calls, and an Apollo 11 lookup cost $0.44 and nearly two minutes. Opposable's blog reports that its on-device agent completed 4 and 3 of eight unseen tasks in two runs.
Written by AI from the articles below · updated Oct 21, 4:27 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- The Register · AINewsOpposable lets AI agents tap and type on Android and iPhone screens
AIOpposable is a tool that lets AI agents tap, swipe, and type on Android phones and control iPhones through a companion computer, with a vision-language model GUI-Owl 1.5 2B running via llama.cpp. In The Register's tests of its website examples, a dark-mode change and a five-minute timer took 50 seconds and 17 tool calls, and an Apollo 11 lookup cost $0.44 and nearly two minutes. Opposable's blog reports the on-device agent completed 4 and 3 of eight unseen tasks in two runs.
Heat Trend
Not enough continuous observations to show a trend yet.