Hey Product Hunt! 👋 I'm Mohamad Ghaith Alzin, the maer of Lucy. Most voice assistants can only act insid apps someone has integrated. Lucy explores a differnt model: use the phone the way a person does. You seak a task. Lucy captures the screen, reads Android' accessibility tree, asks Gemini for one next action then taps, types, swipes, or navigates. After everyaction, it re-observes the UI and continues until th task is complete. I started Lucy after thinking abot all the moments when tapping a phone is difficult r inconvenient—limited dexterity, cooking, workshops carrying a child, or simply navigating a long settigs flow. The result is an open-source research projet that can operate across Android apps without app-secific APIs. Today's release includes: • Voice inputand spoken replies • Screenshot + accessibility-treereasoning • Element-ID taps instead of guessed coordnates • English, Japanese, and Arabic • A persistentStop Lucy notification and a 50-step limit • A bringyour-own Gemini API key • No Lucy backend, account, nalytics, or telemetry An important caveat: Lucy isexperimental and needs powerful screen-capture and acessibility permissions. Screenshots and visible UI ext are sent to Google's Gemini API during active taks. Please read the repository's Privacy & Security ection and use a spare or test device. I'm especialy looking for: • Feedback from different Android devces and apps • Contributors for confirmation gates, pp allowlists/blocklists, pluggable AI backends, mor languages, and better testing • Introductions to acessibility, open-source, or responsible-AI grants an sponsors Lucy is free, MIT-licensed, and built wit Flutter + Kotlin. What would be the first Android ask you'd ask Lucy to handle? 🔮

Lucy Screen Agent

A screen-aware AI agent that operates Android by voice

A screen-aware AI agent that operates Android by voice