FAQ
Plainly.
Practical questions, answered honestly, grouped so you can find yours fast. If something here is vague, tell us — we'd rather fix the answer than leave it soft.
Product
What Nova is.
Does Nova replace ChatGPT?
No — it does a different job, and you'll likely use both. A chat assistant lives in one window and starts fresh each time. Nova runs on your computer: it understands your screen, remembers your work across sessions, and can act across your apps. It answers questions and helps conversationally too — its purpose just extends beyond a standalone chat box.
What actually makes it different?
Memory and reach. Nova keeps a persistent, private record of your work and operates across your whole desktop, not inside one chat box. The full loop is on how it works.
Does Nova work offline?
The core — memory and desktop perception — is designed to run on-device, so the everyday loop doesn't depend on the network. Optional cloud reasoning is the exception, and it's yours to enable.
Local & cloud
Where the AI runs.
Can Nova use local models?
Yes. Nova is local-first by design — on-device models run the core of what it does, so your work can be understood without leaving your machine.
Can Nova use cloud models?
Optionally, and per task. When something genuinely needs more than local reasoning, Nova can call the cloud — but only opt-in, every call is named, and your data is never used to train a model. The data-path is on local + cloud.
Can I bring my own models or API keys?
That's on the roadmap, not shipped yet. Bring-your-own models and keys are part of the model-orchestration direction; we'll mark it clearly the moment it's real.
Privacy & memory
Your data.
Does Nova remember everything?
No. It remembers only what you give it access to, never passwords or secrets, and never anything from apps and files you keep out of scope. What it does keep, you can see — on memory.
Can I delete memories?
Yes — inspect any memory, correct it, or wipe everything with one command. Because the index is local, gone means gone; a reload proves it.
Is my data used to train models?
Never — not local data, not opt-in cloud payloads. It's a line we don't cross, on our side or with the providers a cloud call passes through.
How is my data protected?
On-device by default, encrypted at rest. Your screen and files stay on your machine. Cloud features are opt-in, named, and never retained. The detailed version is on privacy & security.
Platforms & access
Getting it.
What operating systems are supported?
Windows first, with macOS and Linux to follow. Nova is built native on each so it works like part of the system, not a browser tab pretending to be one.
How do I get access?
Join the waitlist on the homepage. Nova is in an invitation-only preview; we onboard people in small groups and write when it’s your turn. Requirements and channels are on get Nova.
How does pricing work?
It isn't set yet. Nova is in an invitation-only preview; pricing will be shared closer to a public launch. We won't surprise preview users with it.
What does joining the waitlist mean?
It puts you in line for early access as we open the preview up. We write to you directly when it's your turn, onboard in small groups, and work closely with early users. No spam, no obligation, and no pricing surprises.
The team
Who's behind it.
Who is building Nova?
A small, independent team building the tool we want to use ourselves. How we think and decide is on about.
See it do real work.
Nova is in private preview. Request access, and we’ll write when it’s your turn.
Request access