Summary
Nimbus8 is built around a simple rule: local AI work should stay local. Chats, prompts, attachments, model weights, search indexes, memory, schedules, agent runs, and most logs live in the app sandbox on your device.
There is no Nimbus account required for the app, no advertising SDK, no tracking SDK, no default product analytics, and no Nimbus-hosted AI inference path for your prompts. Network features exist, but they are opt-in and scoped to the provider or action you choose.
What stays on-device
Nimbus8 stores and processes these categories locally unless you choose a feature that sends them elsewhere:
- Conversations you have with any module (Gale, Cirrus, Ashe, and others).
- Files, images, and attachments you hand to a module.
- Model weights, local model metadata, and device-fit decisions.
- Local search indexes, memory, transcripts, generated media, and exports.
- Agent runs, scheduled Hands, tool history, diffs, and activity feeds.
- Preferences such as theme, language, default models, enabled modules, and safety settings.
Local model inference runs on your Apple device through Apple-native runtimes and open-source engines, including:
- MLX and Apple Silicon runtimes for local open models.
- llama.cpp and GGUF for broader compatibility and quantized models.
- Core ML, Apple Vision, and Apple Speech for supported vision, OCR, and audio tasks.
The network is not required for local chat, local search over imported content, local agent memory, or local model runs after the relevant model is installed.
Accounts, analytics & payments
Nimbus8 does not require a Nimbus account to use the app. We do not ship advertising SDKs, tracking SDKs, or default product analytics. We do not use your local content to profile you, train models, or improve Nimbus8 in aggregate.
Nimbus8 Pro is purchased through Apple In-App Purchase. Apple handles billing, taxes, refunds, family sharing eligibility, and restore state through your Apple ID. Nimbus8 does not see your card number and does not store checkout email, receipt IDs, or bearer license tokens.
Optional network features
Nimbus8 uses the network when you explicitly invoke something that needs it. Examples include:
- Browsing or downloading models from Hugging Face, Civitai, or another model catalog.
- Using your own API key with a cloud LLM provider such as OpenRouter or an OpenAI-compatible endpoint.
- Running Mist web search through a search provider you configure.
- Creating GitHub pull requests, reading repositories you authorize, or connecting another source provider.
- Opening E2B previews, building in a sandbox, or deploying to Vercel, Netlify, Railway, or another deploy account you connect.
- Contacting support or validating App Store entitlement state when required.
When you use those features, Nimbus8 sends only what the feature requires: for example, a model identifier, a search query, selected project files, a PR body, a deployment bundle, or the provider API request. Provider privacy policies apply to those interactions.
Nimbus servers
Nimbus servers exist only for tasks the device cannot fully perform or where a server endpoint is necessary: build/deploy orchestration, support intake, receipt or entitlement plumbing, and security/abuse controls around those endpoints.
Nimbus servers do not run AI inference for your prompts, do not train models on your content, do not maintain an always-on agent daemon, and do not sync your local chats, files, model weights, or agent memory.
Keys & credentials
Bring-your-own keys are stored in Keychain on your device and are used only for the provider you configure. Do not paste keys into prompts. Removing a key or deleting the app removes Nimbus8's local access to that credential.
Some credentials may be synchronized by Apple if you enable iCloud Keychain at the system level. Nimbus8 does not upload those credentials to Nimbus servers.
Retention & deletion
Local data stays on your device until you delete it or uninstall Nimbus8. You can delete chats, models, imported files, generated outputs, local indexes, and agent runs from inside the app where those surfaces expose deletion controls. Uninstalling Nimbus8 removes app-container data from the device according to Apple's platform behavior.
Data sent to Apple or a connected third-party provider follows that provider's retention policy. Support messages you voluntarily send may be kept long enough to respond and maintain security or abuse records.
Security
Nimbus8 relies on the iOS, iPadOS, and macOS sandbox, Keychain, system permission prompts, and encrypted transport for network features. Local files are protected by the passcode, FileVault, Face ID, Touch ID, and other security settings you configure on your Apple device.
Children
Nimbus8 is intended for users aged 13 and over. We do not knowingly collect personal information from children under 13. Because the app does not require a Nimbus account or default telemetry, we do not maintain profiles that identify children or adults.
Changes to this policy
If we materially change this policy, we will update the date at the top and call out the change in the Nimbus8 changelog or release notes.
Contact
Questions about this policy? Email support@driftrail.com or visit the support page.