mincoffical

macOS and Linux

M1

A private AI agent for macOS and Linux. It runs K2 Horizon 3.7B by IFM on your own computer, thinks as hard as you ask, and gets things done without anything leaving it.

Download for Mac Other waysGitHub All projects
Model3.7BK2 Horizon 3.7B, a reasoning model from IFM, answers every message.
Effort levels3Low, Medium or High. Pick how long it thinks right next to the message box.
Cloud calls0Answers come from your computer. The only traffic is downloads and the web searches you ask for.

What it does

An AI agent that lives on your computer, not on someone else's server.

Ask it something, or ask it to do something. It runs commands, reads and writes files, searches the web, sets reminders, uses the clipboard and clicks around websites, and you choose how much it may do without asking first.

Try it withfind the biggest files in Downloadsremind me to stretch at 4plan a week of dinners, effort HighTap the mic and just talk. Parakeet 110M writes it down.
Installing
Terminal
curl -fsSL https://raw.githubusercontent.com/morriszdweck/m1/main/install.sh | bash→ Downloading the app→ Opening the setup screenDownload K2 Horizon 3.7B in the browser, then say hi.
  1. 01

    Get the app

    Open the Mac app, or paste one line on Linux. Either way a ready-to-run copy goes in ~/.m1 with an m1 command. No npm, no build step.

  2. 02

    Pick a size

    The setup screen opens in the app (or at 127.0.0.1:8270). Choose Smallest, Balanced or Best.

  3. 03

    Set the effort

    Click Low, Medium or High beside the message box and slide. Higher means it thinks longer before it answers. Or tap the mic and say it.

  4. 04

    Set the leash

    Ask for Approval, Approve for Me or Full Auto. Every action it takes lands in Activity.

Why M1

Private

The server only listens on this computer and wants a token that changes every launch. Chats, memory and files stay on your disk.

Thinks on a dial

K2 Horizon reasons before it answers. Low is quick, High works it through. The reasoning stays folded away until you tap it.

Safety you pick

Approve for Me uses the small Auto 150M model to approve safe actions and ask you about the rest. If Liquid Agent already has it, M1 uses that copy.

Install it

Download for Mac Intel Mac

Drag M1 to Applications and open it; it sets itself up on first launch. The app isn't notarized yet, so macOS blocks it the first time. Either install with the terminal command below instead, or open System Settings, Privacy & Security, scroll down and click Open Anyway. On Linux, use the terminal command:

$ curl -fsSL https://raw.githubusercontent.com/morriszdweck/m1/main/install.sh | bash
$ m1 update   # update now (it also updates by itself)
$ m1 stop     # stop it

You need

An Apple silicon Mac with macOS 14 or later, or a 64-bit Linux PC from Ubuntu 22.04, Debian 12, Fedora 36 or newer, and about 4 GB free for the Balanced size. On Linux it uses your GPU through Vulkan, or the CPU.

Afterwards

Open it from Applications, Spotlight or your app menu, or run m1. New versions install themselves when nothing is running. Settings, Downloads switches model sizes or any GGUF quant.

From source

Clone the repo and run scripts/start.sh, or run npm ci, npm run build and npm start.