First run
Grux opens on a short setup that asks before it assumes, and every screen in it can be skipped and come back to later.
Rendered unchanged from the Grux README at commit 86963f2, fetched 2026-11-01. If this page and the repository disagree, the repository is right.
- One question: what you want Grux for. The answer picks the features to start with, and the next screen shows the pick before asking for anything, so you can change it. Would rather choose yourself? "I would rather pick from a list" asks how much to set up now instead.
- Your name and a model, then a short look at how Grux works. With Ollama running and no model yet, the model step offers the Grux Local made for this Mac's memory; one press downloads it and points chat at it, and nothing downloads before the press.
- Setup, one thing at a time: only what the picked features need. Every macOS permission gets its own screen that says what it is for before macOS asks.
The minimum useful setup is one model. Chat runs on an Anthropic API key, or on Ollama with no key at all.
If you would rather not send anything to a hosted model at all, install Ollama, pull a model, and turn on Offline mode in Settings, Models: Grux then uses it with no key. Ask Grux to open Local Models to see the models you have and the ones that fit your Mac.
Credentials go to the macOS Keychain under the service com.gruxai.grux. They are never written to disk in plaintext and never leave the machine except to the provider they belong to.
Free, MIT licensed. macOS 14 or later, Apple silicon. 28.4 MB, signed
and notarized by Apple. No account, no server, no subscription.