Features
Run 1000+ local LLMs offline
One-click model download from Hugging Face
Built-in TurboQuant (8x faster inference, 6x less memory)
Chat with persistent memory across sessions
Project organization for context switching
OpenAI-compatible local API endpoint for agents
Document chat (RAG) support
Code review and refactoring
Private document analysis
GGUF, MLX, ONNX model format support
No account or sign-up required
Open-source under Apache-2.0
Available on macOS, Windows, Linux, iOS, Android
Supports agents: Hermes, OpenClaw, Cline
TurboQuant: zero accuracy loss at 3 bits
Multimodal input: text, images, audio, video
Real-time Google Search integration
Computer-use actions (clicking, typing) in Gemini 3.5 Flash
Large context window for long documents
Google service integration: Gmail, Docs, Maps, Calendar, Drive, YouTube
File upload support: PDF, images, code files
Code generation, explanation, and debugging
Creative writing assistance
Web and mobile access (Android/iOS)
Google Home Speaker integration
Apple AI architecture integration via partnership
Direct interface control in Gemini 3.5 Flash
Real-time outage monitoring (Hacker News reports)
