About Kynora

AI that works
for you, not on you.

Kynora is a private knowledge workspace built on a simple belief: your documents, queries, and conversations belong to you.

Everything you need to know

0
Cloud API calls made during a session
100%
Of inference runs on your own hardware
3
Platforms — macOS, iPadOS, iPhone

Product

Category AI Productivity App
Platforms macOS · iPadOS · iOS
Distribution App Store
Status Coming Soon
Pricing Waitlist (TBD)

Technology

AI Runtime MLX · llama.cpp
LLM Models Open Source Models
File Formats PDF · DOCX · TXT · Markdown
Data Storage On-device only
Telemetry None

The problem with cloud AI tools

Most cloud AI tools operate the same way: you upload documents, servers process documents. For many tasks, that trade-off is acceptable. For professionals working with sensitive material, it isn't.

Lawyers can't upload client contracts. Doctors can't process patient notes. Researchers can't submit unpublished data. Executives can't analyse board materials. The tools that would help the most are the ones that simply can't be used.

Kynora is the answer to that gap. It runs the full AI pipeline — document parsing, text embedding, vector search, and language model inference — entirely on your device. MLX makes this fast enough to be genuinely useful. Kynora makes it simple enough for anyone.

There is no backend server. No account. No subscription that requires sending data anywhere. Your knowledge base lives in your ~/Library, under your control, forever.

Privacy by architecture
No backend = nothing to leak
Offline-first, always
Works on a plane, in a cabin, anywhere
Open models, your choice
Qwen, Gemma, Llama — no lock-in
Cited, verifiable answers
Every answer traces back to a source
Native performance
MLX & llama.cpp

Built by one person who got tired of the trade-off

Sagar Purohit
Developer

Kynora is a solo indie project, built from scratch in Swift 6 for macOS, iPadOS, and iPhone. The idea came from a simple frustration: the best AI tools for knowledge work all require handing over the exact documents you most need to keep private.

The answer wasn't to build a better privacy policy. It was to build an app with no server to send data to. All AI inference, embedding, vector search, and storage runs locally, using Apple's own frameworks and open-source models. The app doesn't even have a login screen.

Kynora is currently in development. If this resonates with you — anyone who works with sensitive documents — join the waitlist and you'll be the first to know when it ships.

Be first to know when Kynora launches

Join the waitlist. No spam — just a single email when it's ready.

801+ people already joined