A thought arrives. You catch it. Niva helps it find its place — without asking you to organize your mind before you can write it down.
VOICE-FIRST · EDITABLE · MADE FOR REAL LIFE
Niva is being built around a simple idea: capture should feel effortless, and organization should happen in the background — with you still in charge.
Catch a fleeting thought with your voice, or write and paste a note the way you already do.
Tasks, reminders, ideas, and reference notes can take different paths, so a to-do does not disappear into a pile of thoughts.
Related notes gather around folders and appear in a visual map. Strong matches can connect automatically; uncertain ones stay yours to decide.
Long thoughts deserve room. Niva is designed to keep typed and pasted writing editable as one note, save drafts locally as you work, and send changes in the background.
Niva keeps the capture experience close to the person and its decision engine behind a server boundary. That separation lets us test models without putting provider keys or model policy in the mobile app.
The iOS experience is being built in SwiftUI. It brings together capture, an editable note workspace, folders, a visual knowledge map, and review.
Niva requests on-device speech recognition when the iOS device and selected language support it. Niva does not save audio recordings. Draft text is protected locally while you write, then saved in the background.
Firebase Authentication supports the app session. Firebase SQL Connect provides the typed data interface to PostgreSQL for notes, folders, and relationships.
Firebase Cloud Functions run Niva's server-side decision flow. Jev is the structured classification and alignment layer; the current model provider path is OpenRouter. Secrets stay off-device.
Zoom out to see the landscape. Folders become landmarks; their notes gather nearby; confirmed links show where one thought leads to another. Tap a note to return to the words themselves.
Niva is being built to make organization feel lighter — not to make decisions about your life on your behalf.
A suggestion is not a replacement for your note. You can open and change the original text.
Strong matches can be linked automatically. Less certain suggestions are kept distinct so you can decide.
Niva requests on-device speech recognition where iOS supports it and does not save audio recordings.
Niva's decision layer helps classify and align notes. Product rules and your edits remain in control.
No. Niva's current server-side model path uses Jev through OpenRouter. We plan to evaluate the first-party Claude API as a possible model option; we will only describe it as integrated if and when that work is actually shipped.
Jev is Niva's structured decision layer for classifying notes and evaluating possible relationships. It is an internal part of Niva, not a separate note-taking product.
Niva does not save audio recordings. It requests on-device speech recognition where the iOS device and selected language support it. The resulting text can be reviewed and edited as a note.
Niva is preparing for a closed beta.
Back to the beginning