Kynora is a private, offline-first AI workspace for macOS, iPadOS, and iPhone. You import documents — PDFs, Word files, text files, and notes — and Kynora builds a searchable AI knowledge base entirely on your device. No cloud servers, no accounts, and no internet connection required.
Yes, completely. Kynora runs all AI inference, text embedding, and vector search locally on your device using Apple's MLX framework on Apple Silicon or llama.cpp on other hardware. Once the app and models are installed, it operates entirely offline — even in airplane mode.
Cloud-based AI tools send your documents to external servers for processing. Kynora does the opposite — everything runs locally on your device. Your files, queries, and conversation history never leave your machine. There are no accounts, no subscriptions, and no third party that can access your data.
Kynora supports PDF, DOCX, TXT, and Markdown files. Scanned PDFs and images are handled via on-device OCR using Apple's Vision framework. Additional formats are planned for future releases.
Kynora uses open-source local LLMs including Qwen 3B, 7B, and 14B, Gemma 4B, and Llama 3.2 variants — all running on your device. For document embeddings it uses BGE-Small-EN and Nomic Embed Text. You choose which models to download; nothing large is bundled by default.
Kynora supports macOS 14 (Sonoma) and later, iPadOS 17 and later, and iOS 17 and later. Apple Silicon (M1 or later) is recommended for the best on-device AI performance, but Intel Macs are also supported via llama.cpp.
Kynora has no backend server — there is nothing to send your data to. All processing happens locally: document parsing, text chunking, AI embedding, vector search, and LLM inference all run on your device. Even model downloads go directly from Hugging Face to your device, bypassing any Kynora server entirely.
Kynora is currently in development. Join the waitlist to be notified at launch and to receive early access details. Pricing has not been announced yet.