How Kynora AI works,
step by step

From opening the app for the first time to getting a cited answer from your own documents — here's the full walkthrough of how Kynora AI works.

Four steps to your first answer

STEP 01

Install & open

No account, no sign-in. Pick a plan during setup — monthly and yearly start with a 7-day free trial.

STEP 02

Choose your models

Download the recommended model for your device during setup; add an embedding model when you start with Ask Docs.

STEP 03

Chat or RAG (Ask Docs)

Talk to your local model, or import documents for cited, grounded answers.

STEP 04

Work offline, always

Once models are downloaded, everything runs without an internet connection.

See every step in action

From first launch to cited answers and Projects — the complete flow on Mac and iPhone. Jump straight to any part with the chapters below.

Every step, in detail

What each part of Kynora AI does, and how the pieces fit together.

📥

1. Install and Open Kynora AI

Open Kynora AI on your Mac, iPad, or iPhone and start immediately — there's no account to create and no sign-in screen. There's no Kynora AI user ID and no server-side profile; everything you do lives in the app's own local storage on your device. The last step of setup is choosing a plan: start a 7-day free trial on a monthly ($4.99) or yearly ($24.99) plan, or pay $49.99 once for Lifetime. Purchases go through your Apple Account and the App Store — Kynora AI never sees your payment details.

⚙️

2. Choose Your Models

On first launch, you tell Kynora AI what you'll mostly use it for, and it recommends a generation model (the model that writes answers) that fits your Mac, iPad, or iPhone's memory and chip. For RAG (Ask Docs) you'll also need an embedding model (the model that indexes your documents for search) — pick one in Settings → Ask Docs or from Models. Kynora AI downloads whatever you pick directly from that model's host — for example Hugging Face — straight to your device. Kynora AI doesn't sit in the middle of that download or see its contents. Apple Silicon devices run both MLX and GGUF models; Intel Macs run GGUF models through llama.cpp.

💬

3. Start a Chat

Use Chat for everyday questions, drafting, and brainstorming with your local AI model — no documents required. Attach images for models that support vision, use on-device speech dictation instead of typing, and pin a specific model to an individual conversation if you want a different model for that thread. Tap the speaker under any answer to hear it read aloud, or turn on voice mode in Settings → General for a hands-free conversation. On a Mac, press ⌃⌥J in any app to open Quick Ask, a floating box that asks your local AI without leaving what you're doing.

📚

4. Build a Knowledge Base with Ask Docs

Import the files you already have — PDF, DOCX, PPTX, XLSX, HTML, TXT, Markdown, and images are all supported, as long as an embedding model is set in Settings → Ask Docs. Kynora AI parses each file, splits it into chunks, and embeds those chunks into a private, on-device index using hybrid retrieval — dense vector search plus keyword search, with an optional reranker for extra precision — all without the files ever leaving your device.

🔎

5. Ask Questions and Verify the Citations

Open a document in Ask Docs and ask a plain-language question — Kynora AI retrieves the most relevant passages from that document, then generates an answer grounded in them. Every answer lists its sources — document title, section heading, and page label (on Mac, hover a source to see the excerpt) — so you can check the original yourself instead of taking the answer on faith. See our explainer on how RAG works for the mechanics behind this.

🖨️

6. Search Scanned Documents with OCR

Scanned PDFs and photographed documents just work: when a page has no text layer, or you import an image file (JPG, PNG, HEIC, and more), Apple's Vision framework reads the text on-device, so scanned pages become searchable the same way a native PDF would be. To also pull text out of pictures embedded in PDF, DOCX, and PPTX files, turn on Extract text from images in Settings → Ask Docs — it's off by default.

🗂️

7. Organize with Projects and Archive

Group related conversations, including Ask Docs chats, into a Project, and Kynora AI generates a project-level summary as it grows. Move conversations you're done with into Archive to keep your active workspace focused, without deleting anything.

🔄

8. Manage Your Models Over Time

Swap your generation or embedding model whenever you want — switching embedding models means re-indexing your existing documents against the new one. On a Mac, advanced users can also point Chat or Ask Docs at a local Ollama server instead of the built-in models (Settings → API); it's off by default. If you turn it on, conversation text, including any document excerpts Ask Docs retrieves, is sent to that server.

🔒

9. Lock, Back Up, and Make It Yours

In Settings → General, turn on App Lock to require Face ID, Touch ID, or your passcode to open Kynora AI. Because nothing lives in a cloud, use Backup to save your chats, projects, and preferences to a file you keep, and restore it on a new device. You can also pick light or dark mode, an accent color, your chat font and text size, the send key, the export format, and optional automatic cleanup of old chats.

✈️

10. Go Offline, Whenever You Want

Once your chosen models are downloaded, Kynora AI needs no internet connection at all — airplane mode, no Wi-Fi, a remote location, it all works identically, because parsing, embedding, retrieval, and generation all run on your device.

Disclaimer: Kynora AI orchestrates retrieval and prompting, but the open-source model you chose in Step 2 is what actually writes each response — Kynora AI facilitates the conversation between that model and you, it doesn't author the answer. AI-generated output can be incomplete or wrong, so verify anything important against the original document or source. See our Privacy Policy and Terms of Service for the full picture.

Keep exploring