How Cloaked Protects Your Conversations
Not "we won't read your data." We can't. Here's exactly how that works.
On-Device Architecture
You type a prompt
Stays in app memory
MLX processes on-chip
Apple Silicon neural engine
Response generated locally
Nothing leaves your device
The Five Zeros
Zero Network Calls
After downloading a model, Cloaked never contacts any server. Your prompts and responses exist only in your device's memory.
Zero Analytics
No usage tracking, no crash reporters, no telemetry of any kind. We literally cannot see how you use the app.
Zero Accounts
No sign-up, no email, no phone number. Download and start chatting. That's it.
Zero Cloud Storage
All conversations, projects, and memory data are stored in the app's local sandbox. Not iCloud. Not our servers. Nowhere.
Zero Third-Party Access
No SDK shares data with advertisers, analytics platforms, or social networks. The app contains zero third-party code that phones home.
Technical Foundation
Cloaked uses Apple's MLX framework to run language models directly on your device's Apple Silicon chip. MLX is optimized for the unified memory architecture of M-series and A-series processors, delivering efficient inference without requiring a GPU server.
Model files are stored in the app's sandboxed container — a protected area on your device that other apps cannot access. Conversation data is persisted using Apple's SwiftData framework with CloudKit explicitly disabled.
All models use permissive open-source licenses (Apache 2.0, MIT, or similar). You can verify the model source code on Hugging Face. We don't train, fine-tune, or modify any model with your data.
What About Web Search?
When you enable web search, Cloaked sends your query to DuckDuckGo's API — a search engine that doesn't track users or store search histories. The search results are fetched, processed entirely on your device, and fed to the local model for a contextual response. No search data is stored after the session.
Experience truly private AI.