Fae is in early development. Here's what you'll need and how to get set up.
Apple Silicon (M1 or later) recommended. macOS 14+ for a pure native SwiftUI experience — Metal-powered orb animation, native conversation view, and native settings. No web views, no Electron. Feels like it belongs on your Mac from the moment you open it.
x86_64 with CUDA-capable GPU for local inference. Headless fae-host binary with JSON protocol — connect any frontend.
x86_64 with CUDA GPU support. Same headless fae-host.exe bridge via named pipes. Connect your preferred UI.
Local AI model selection adapts to your available RAM:
| System RAM | Model | Capabilities |
|---|---|---|
| 24 GiB+ | Qwen3-VL-8B-Instruct | Vision + text, stronger tool calling and coding |
| < 24 GiB | Qwen3-VL-4B-Instruct | Vision + text, lighter footprint |
Grab the latest release for your platform from the GitHub releases page. On macOS, download the .dmg and drag Fae to your Applications folder. On Linux and Windows, extract the archive to your preferred location.
On first launch, Fae will guide you through granting the permissions she needs — microphone for listening, and optionally camera, contacts, calendar, and reminders depending on how you want to use her. You can change these at any time in Settings.
Fae's onboarding is a gentle conversation. She'll introduce herself, learn your name (from your Me Card on macOS, or by asking), and get a sense of how you'd like to work together. No forms, no setup wizards — just a chat.
That's it. Fae is always listening (unless you tell her to sleep). Talk to her naturally, ask for help, set reminders, or just have a conversation. She'll learn about you over time and get better at anticipating what you need.
We're in active early development. Sign up for the waitlist and we'll let you know the moment Fae is ready to come home.
Join the Waitlist