JEV ADB

•

Control and automate Android devices over USB or wireless ADB using fast decision models, typed UI grounding, and built-in safety controls.

JEV ADB connects AI agents directly to Android devices over USB or wireless debugging to enable autonomous mobile workflows and direct UI interactions. By transforming raw uiautomator dumps into compact, typed element models and leveraging fast decision models like TypeSafe's JEV or OpenRouter's Mercury, it enables agentic phone navigation in ~100 ms per step. Built with a local-first architecture, the tool emphasizes device security with automatic password masking, strict safety gates against destructive operations, and an explicit consent model that requires manual device authorization before any actuating commands are run.

Key Features

010 GitHub stars
02Comprehensive connectivity supporting both wired USB and Android 11+ wireless pairing
03Typed UI parsing that transforms uiautomator dumps into compact trees while masking sensitive fields
04Direct device control for tapping, swiping, text entry, key events, and app launches by stable references
05Autonomous goal-driven task loop with step and USD budget boundaries
06Strict safety architecture featuring persistent device consent and safety gates for irreversible actions

Use Cases

01Automating multi-step workflows, settings adjustments, and daily routines across Android apps
02Executing end-to-end mobile QA testing and autonomous user-flow verification on physical devices
03Enabling AI coding assistants to inspect, interact with, and troubleshoot running mobile applications