LUI replaces your Android home screen with a single conversational interface. Voice or text — 87 tools execute your intent instantly.
Wake word ("Hey LUI"), conversation mode, streaming TTS. Works hands-free, even when the phone is locked.
Qwen 3.5 0.8B runs locally. No cloud, no API keys, fully private. Auto-downloads from the Connection Hub.
Gemini, Claude, OpenAI, or Ollama with native structured function calling. 5-round tool chaining.
Search DuckDuckGo, browse any URL. Results shown as clickable rich cards. No API key needed.
Camera capture and gallery picker with AI image analysis. "Take a photo and tell me what you see."
Geofencing and scheduled actions. "When I leave home, enable DND." "In 5 minutes, lock my phone."
Notification triage. Urgent apps pass through, noise batched into an Evening Digest. 2FA codes auto-extracted.
Read any app's screen, tap buttons, type text, scroll. The LLM can drive any app on your phone.
WebSocket server for remote AI agents. pip install lui-bridge. Event streaming, bidirectional, permission tiers.
Stop texting sandboxed bots. Connect any MCP-compatible agent and give it first-class access to Android hardware, sensors, notifications, and screen control.
Full conversation + command execution. Pipe tasks through Claude's CLI directly from your phone.
Native MCP via MCPorter. Connects directly to LUI's bridge — zero adapter code.
Single-query agent execution. Optimized for fast one-off tasks via the -q flag.
pip install lui-bridge. 10 lines of Python. Any framework, any model, full phone control.
Clone, build, install. Or connect your agent to an existing LUI device.
Open source. GPLv3. Made for builders.