Summary
Nimbus8 is built so that the data you put into it never needs to leave your device. Inference happens locally. Chats, files, and prompts are stored in the app's iOS sandbox. There is no account, no cloud sync, and no telemetry by default.
This page spells out the specifics: what data exists, where it sits, what leaves the device (and when), and the controls you have over all of it.
What we collect
Nimbus8 stores the following on your device only:
- Conversations you have with any module (Gale, Cirrus, Ashe, and others).
- Files, images, and attachments you hand to a module.
- Model weights you download, and metadata about them.
- Agent runs, scheduled hands, and their activity feed.
- Preferences — theme, default models, enabled modules.
We do not maintain an account for you. We do not have a server that holds any of the above.
How we use it
All of the data listed above is used exclusively to let the app work — to run your next inference, load your last chat, keep your scheduled agent on track. None of it is used to profile you, train models, or improve Nimbus8 in aggregate.
On-device processing
Every model Nimbus8 ships with runs locally, on your iPhone or iPad, using Apple-native runtimes:
- MLX on Apple Silicon for the flagship open models.
- llama.cpp + GGUF for broader compatibility and smaller quantizations.
- Core ML for vision, OCR, and audio front-ends.
A capability manifest routes each prompt to the engine and model best suited to your hardware. The network is not required for any of this.
Network & optional services
Nimbus8 uses the network only when you explicitly invoke something that needs it. The default posture is offline. Examples of opt-in network use:
- Browsing or downloading a model from Hugging Face.
- Fetching a favicon for a URL a module renders.
- Publishing a pull request from Cirrus to your GitHub account, with your explicit confirmation each time.
- Optional web search in Mist, which is clearly toggled and visually distinct.
When you do use these features, only the minimum data required for that feature is sent — a search query, a model identifier, a PR body — and nothing else from your device goes along with it.
Third parties
Nimbus8 does not integrate analytics SDKs, advertising SDKs, or crash-reporting SDKs. When a feature you invoke talks to a third party (for example, downloading a model from Hugging Face), that third party's privacy policy applies to that interaction.
Children
Nimbus8 is intended for users aged 13 and over. We do not knowingly collect data from children under 13. Because Nimbus8 stores data only on-device and has no account system, we have no mechanism to identify or profile any individual user — under 13 or otherwise.
Retention & deletion
Your data stays on your device for as long as you keep it there. You can clear any conversation, uninstall a model, or delete an agent run from within the app. Uninstalling Nimbus8 removes all of it — including every stored model — from your device.
Your rights
Because your data does not leave your device, traditional data-subject requests (access, deletion, export) are satisfied directly in the app: you already have full access, you can delete anything, and export is available as plain text for conversations and standard formats for files.
Security
Nimbus8 relies on the iOS sandbox and Keychain for its security boundary. App data is protected by the device passcode and Face ID / Touch ID where you've configured them. The app does not bypass system protections, does not require unusual entitlements, and never elevates privileges.
Changes to this policy
If we materially change this policy, we'll update the date at the top and call out the change in the Nimbus8 changelog. We don't have a mailing list to notify you, because we don't have your email.
Contact
Questions about this policy? Reach the team through the community page or the support link in the app's settings. We try to reply within a few days.