Skip to content

Documentation

Elythros runs a language model on your iPhone. Your conversations, the documents you add, and anything you dictate stay in the app's own storage on the device.

Two features can send something elsewhere. Both are off until you turn them on, and the app tells you what each one sends at the point where you enable it:

  • Web search sends a derived query — never your conversation — to a provider you supply your own key for.
  • Mac offload sends your prompt to a Mac you have paired, so it can answer with a model far larger than a phone can hold.

Everything else happens on the device, including generating replies, reading and indexing your documents, recognising text in scanned pages, transcribing your voice, and speaking replies aloud.

Getting started

  1. Open the app. Apple's built-in system model needs no download and is ready immediately, so you can ask something straight away.
  2. Pick a model that suits your phone. The built-in one is a floor, not a ceiling — see choosing a model for what fits and what each is good at.
  3. Add your documents if you want answers grounded in your own files. They are read and indexed on the device, and answers cite the page they came from.
  4. Pair a Mac if you have one, to use a model no phone could hold.

What you need

iPhoneiOS 26 or later
MacmacOS 26 or later on Apple silicon — only for offload

Running a model on the phone wants recent hardware and storage to spare. The app shows you the download size and the memory a model needs before it downloads anything.

There is no account

There is nothing to sign up for, no profile, and no server of ours that the app talks to. That is also why some things are yours to arrange rather than ours to provide — reaching your Mac from outside your own network being the main one.

Where to go next