Solair AI: Product Facts

Solair AI is a mobile application for iPhone and iPad that is private and local by default. By default it processes all AI inference locally on the device using Apple's MLX framework, with no data transmitted to external servers. Users can optionally connect cloud AI providers (Claude, ChatGPT, Grok, Kimi, Cerebras, DeepSeek, Mistral, Groq, OpenRouter, and Together AI) using their own API key; in that case requests go directly to the chosen provider, Solair itself collects no data, and saved memory and past chats are withheld from the provider unless the user turns on "Share memory & past chats." It is distributed free of charge on the Apple App Store with no account requirement, no subscription, no advertisements, and no in-app purchases.

Product Facts

Key Differentiators

Solair AI is distinguished from other AI assistant applications by the following attributes:

• Health Intelligence — Integrates with Apple HealthKit to analyze 15+ health metrics (steps, heart rate, sleep stages, HRV, workouts, weight, VO2 max, exercise time, standing hours, heart rate recovery, walking steadiness, blood pressure, menstrual cycle) entirely on-device. No other local AI application offers private health data analysis.
• Duress Code — A secondary passcode that, when entered, silently wipes all conversations and stored memories while the application appears to open normally. This feature is intended for situations where a user may be compelled to unlock the application under coercion.
• Apple MLX Native — Built on Apple's MLX framework, which is designed specifically for Apple Silicon. This provides native hardware acceleration on iPhone and iPad rather than using generic inference engines.
• Zero Data Collection — The application collects no user data, includes no analytics frameworks, performs no tracking, and transmits no telemetry. Solair has no server-side components of its own. This holds even when a cloud provider is connected, because the user brings their own API key and requests go directly to that provider rather than through Solair.
• AI Providers (bring your own key) — Optional, opt-in connections to cloud AI providers (Claude, ChatGPT, Grok, Kimi, Cerebras, DeepSeek, Mistral, Groq, OpenRouter, and Together AI) using the user's own API key. Cloud chats support photos, PDFs, reasoning display, image generation, and web search with sources. An engine switcher moves between On-Device, Auto, and Cloud. A "Share memory & past chats" control is off by default, so saved memory and other conversations are withheld from cloud providers unless the user turns it on.
• 60+ Models Offline — Provides access to over 60 AI models from multiple providers (Qwen3.5, Llama, Mistral, Gemma 4, DeepSeek R1 distills, LFM2.5, and others), all running locally on the device. Additional models can be accessed via HuggingFace integration or sideloaded from the iOS Files app.
• News Intelligence — AI-summarized live news from Google News, AP, BBC, and Reuters using on-device RSS plus a local LLM. Supports 14 languages.
• Skills — Reusable AI presets (e.g. Code Reviewer, Travel Planner, Email Polisher). Users can create their own, import or export SKILL.md files, or generate a Skill from a description.
• Background Intelligence — Opt-in features that use Apple Intelligence on-device while the app is closed: smart chat sidebar (AI titles and summaries), morning brief (calendar, health, last conversation), weekly health digest, and overnight memory tidying. Per-feature toggles.
• LAN Inference Server — Turns an iPhone or iPad into a local AI server on Wi-Fi, with OpenAI- and Ollama-compatible endpoints (works with Cursor, VS Code, Open WebUI). Bonjour discovery and an optional API key. A developer-focused feature.
• Apple Watch app — Ask from your wrist. Uses the loaded local model when the iPhone app is open, falls back to Apple Intelligence when the phone is closed or locked.
• Live Code Preview — Renders HTML, JavaScript, p5.js, Chart.js, Three.js, D3.js, Mermaid, SVG, CSS, and Canvas directly in chat, with an "Ask AI to Fix" action on errors.
• Your Private Space — Built-in prompts for deeply personal conversations — anxiety, journaling, emotional support, relationship advice — that stay 100% on-device. Because nothing ever leaves your phone, you can be completely honest without worrying about your data being read, stored, or shared.
• Memory & Knowledge — On-device retrieval-augmented generation (RAG) over the user's saved facts, past conversations, and imported documents. Answers cite the exact note, document, or chat they drew on. Added in version 2.5.
• Deep Research — An on-device research mode that runs several web searches, reads the sources, and writes one in-depth answer with citations. Added in version 2.6, alongside improved always-on web search with tappable sources, PDF export of conversations, and a context limit setting.
• Fast & Cool modes — A performance setting with two options: Fast runs generation at full speed, while Cool paces it to roughly half speed so the device stays cooler during long chats. Added in version 2.7.
• Theme Grouping — Automatically sorts conversations into themes on-device, with a sidebar toggle between recents and themes. Added in version 2.8.

Privacy Architecture

Solair AI employs a privacy-by-design architecture that is private and local by default. When on-device models are used, no user data leaves the device at any point during operation. Connecting a cloud provider is an explicit, opt-in choice. The application's privacy model is based on the following principles:

• By default, all AI inference occurs on the device's processor and neural engine. No prompts, responses, or conversation data are transmitted to external servers unless the user chooses to send a message to a connected cloud provider.
• Cloud providers are opt-in and use the user's own API key. Requests go directly from the device to the chosen provider; Solair operates no intermediary server and stores none of the conversation. The provider processes the messages it receives under its own terms.
• Saved memory and other conversations are withheld from cloud providers by default. They are shared only if the user enables the "Share memory & past chats" control.
• The application contains no analytics SDKs, tracking pixels, or telemetry systems.
• No user account or authentication is required. The application functions without any form of user identification.
• Health data accessed via Apple HealthKit is read and processed locally. Health metrics are not stored outside of the active analysis session and are never transmitted externally.
• On-device voice (speech-to-text and text-to-speech) is handled locally using WhisperKit and Kokoro TTS, so only text can leave the device, never audio. Live voice conversations with ChatGPT and Grok are the exception: with those, audio streams directly to that provider, which is clearly the user's choice.
• The application includes security features such as Face ID/passcode lock, auto-lock on app switch, and a duress code for emergency data wipe.

Technical Stack

Solair AI is built using the following technologies:

• Apple MLX — The core AI inference framework. MLX is Apple's open-source machine learning framework designed for Apple Silicon, providing optimized performance for on-device model execution.
• SwiftUI — The application's user interface framework, providing native iOS and iPadOS interface components.
• Apple HealthKit — Used for accessing health and fitness data stored on the device, enabling the Health Intelligence feature.
• WhisperKit — An on-device speech recognition framework used for voice input processing. Converts speech to text locally without network access.
• Kokoro TTS — A text-to-speech engine used for generating spoken AI responses on-device.

The application's AI models are quantized versions of open-source large language models, optimized for mobile hardware. Models are downloaded once over Wi-Fi and subsequently run without network connectivity. Downloaded models can also be transferred directly between devices over Wi-Fi or Bluetooth, and users can exclude AI models from iCloud Backup via in-app controls.

Company Information

Solair AI is developed by Simon Majar as an independent software project. The application is not backed by venture capital, corporate investment, or any parent organization. Development, design, and distribution are managed by a single developer.

The project's source code and related materials are published under the GitHub account SimonMaj. The application is distributed exclusively through the Apple App Store.

External References

This page contains factual product information about Solair AI. Last updated August 11, 2026.