Fae is a personal AI who runs entirely on your computer. She learns about you over time — your name, your preferences, your world — and keeps everything private and secure. Like having a knowledgeable friend who is always in the room.
Created by Saorsa Labs for Autonomi
In Scottish folklore, the Sìdhe were guardians — watching over people, guiding them home, asking only for respect in return. Fae carries that spirit into the digital world. She handles the complexity — setting up software, managing files, scheduling reminders, researching topics, and keeping track of the people and events that matter to you. You just talk to her.
Fae listens continuously and decides when to speak. She responds to direct conversation, can helpfully interject when she overhears something useful, and stays quiet when it's not her place.
A durable memory system learns your name, preferences, work context, relationships, and routines. All stored locally on your machine. Fae recalls what matters before every response.
Fae extracts dates, people, and commitments from your conversations. She creates reminders, researches topics overnight, and delivers warm morning briefings — all without being asked.
Open apps, manage files, execute commands, and configure software — all through natural conversation. A safety-first approval model keeps you in control of what Fae can do.
With camera permission, Fae can see and understand the world around you. Read text from images, identify objects, and process visual information — all locally and privately.
A four-layer security architecture keeps your data safe. An immutable safety core protects permissions, secrets, and memory — even from Fae herself. Everything stays on your machine. No cloud. No telemetry.
Fae can build new skills based on what you actually need — then test, validate, and install them herself. If anything goes wrong, her built-in Doctor diagnoses and rolls back automatically.
Fae's intelligence runs on a Rust core library, and on macOS the entire interface is pure SwiftUI with a Metal-powered orb — no Electron, no embedded browsers, no web views. On Linux and Windows, a headless bridge exposes the same core via JSON protocol. It feels like it belongs on your machine because it does.
Every conversation, every memory, every piece of intelligence stays on your hardware. Fae doesn't phone home. There's no telemetry, no analytics, no data collection of any kind. Your secrets stay yours.
Qwen3-VL runs on-device via Metal (Mac) or CUDA (Linux). No API keys, no cloud inference.
No usage tracking, no crash reports sent anywhere, no analytics. Fae respects your attention.
All memories are stored as inspectable files on your machine. Ask Fae to forget anything, anytime.
“Fae is calm, practical, and warm. She is an always-present companion, not a summoned servant. Silence is a form of respect and attentiveness.”
— from SOUL.md, Fae's behavioural contract
Most AI assistants appear when summoned and vanish when dismissed. Fae is different. She is always present — like a friend in the room. She decides when to speak and when to stay quiet. She'll interject warmly if she hears something she can help with, and she'll hold her tongue when it's not her place.
Her personality is defined by an open, editable SOUL document that you can customise to match your preferences. She's direct, practical, and honest — she'll never just agree to keep the peace.
Fae is in active early development. Sign up for early access and we'll let you know when she's ready to come home.