# Get started > Install jj-musubi, start a hub, and share a repository between your machines. Documentation for jj-musubi. [Canonical HTML][jj-musubi-canonical]. Source: [`README.md`][jj-musubi-source]. For exact syntax and behavior of the installed version, use `musubi --help`. [jj-musubi-canonical]: https://www.serpentine.com/software/jj-musubi/get-started/ [jj-musubi-source]: https://github.com/bos/jj-musubi/blob/main/README.md ## Install Install musubi on every machine, the hub's included. Download the archive for your platform from the [latest release](https://github.com/bos/jj-musubi/releases/latest) and put the `musubi` in it on your `PATH`. There are builds for Linux (static), macOS, and Windows, each for x86_64 and aarch64. The release's `SHA256SUMS` lists each archive's checksum, and `gh attestation verify --repo bos/jj-musubi` checks that this repository's release workflow built it. On macOS, `musubi` isn't notarized, so macOS refuses to run it if a browser downloaded it. Download it with `curl -LO` instead, or clear the browser's mark with `xattr -d com.apple.quarantine musubi`. Or build it from source, with Rust 1.97.1 or later: ```sh cargo install --locked jj-musubi ``` Or build the latest code, which may not talk to a hub running a release: ```sh cargo install --locked --git https://github.com/bos/jj-musubi jj-musubi ``` To complete musubi's commands and hive names in your shell, see [Complete commands in your shell](reference/commands.md#complete-commands-in-your-shell). ## Quick start One machine that is usually on runs a **hub**. Each repository it shares is a **hive**, and every machine's copy of the repository joins it. A **daemon** on each machine keeps its copies in sync with the hub. You need: - `jj` 0.46.0 and `git` on every machine; - a repository with at least one commit and all of its history: not a shallow or partial clone; - ssh access to the hub's machine from every other machine: if `ssh devbox` works, so does musubi; - `musubi` on the `PATH` of non-interactive shells on the hub's machine, since other machines run it over ssh: check with `ssh devbox musubi --version`. On the hub's machine: ```sh musubi hub ``` It creates `~/musubi-hub` the first time, and prints what to run on the other machines: ```text Created a new hub in /home/you/musubi-hub To share a repository: musubi create --hub devbox To add another machine's copy of it: musubi join --hub devbox To make a new copy of a hive: musubi clone --hub devbox Other machines reach this hub over ssh: use the name they ssh to this machine by, if it isn't devbox. On this machine, leave out --hub. ``` In one machine's copy of the repository, start a hive from it: ```sh musubi create --hub devbox ``` The hive is named after the repository's `origin` remote (`jj-musubi`, say), or its directory. In each other machine's copy, join it: ```sh musubi join --hub devbox ``` The hub works out which hive a copy belongs in from its history and remotes. `create` and `join` start the daemon and remember the hub, so later commands need no `--hub`. The first time a machine uses a hub, musubi adds an ssh key of its own to `~/.ssh/authorized_keys` on the hub's machine, allowed only to reach the hub; ssh may ask for your password that once. To start a new machine from a hive instead: ```sh musubi clone jj-musubi ~/src/jj-musubi --hub devbox ``` To keep the daemon running across reboots, run `musubi daemon install` on each machine. [Set up a hub and machines](guides/set-up.md) covers each step in more depth. ## Day to day Keep using jj as you always have; there's nothing to push, fetch or remember. `jj log` shows each other machine's working copy as a workspace named after it, such as `laptop/default@`, which you can diff, build on, or rebase like any other revision. musubi's own commands are for checking on it: - `musubi` (or `musubi list`) shows the hub's hives, this machine's copies of them, and whether each is syncing. - `musubi status` shows the current repository's hive, whether it syncs and whether the hub has its latest work, every machine's working copies, and the daemon, and warns of what's wrong. It answers at once, for scripts and agents to ask. - `musubi top` shows the hub's traffic and every machine's syncing, live. - `musubi sync` syncs the current repository now; `musubi leave` takes it out of its hive. - `musubi forget ` forgets a machine that went away without leaving. A copy that a coding agent (Claude Code or Codex) joined is forgotten by itself after 6 hours unused (`musubi.forget-after`). - `musubi daemon install` starts the daemon at login; `musubi daemon start`, `stop` and `restart` control it. - `musubi hub install` runs the hub at login; `musubi hub rename` and `musubi hub remove` manage the hub's hives. The daemon logs to `daemon.log` in musubi's state directory (`~/.local/state/musubi` on Linux and macOS), or under systemd, to `journalctl --user -u musubi`. See [Logging](reference/configuration.md#logging). ## Upgrading Upgrade the hub's machine and every other machine together: until 1.0, a release may change the protocol, and a machine and a hub whose protocols differ refuse to sync, saying so. [Upgrade musubi](guides/manage.md#upgrade-musubi) gives the steps, and [CHANGELOG.md](changelog.md) what each release changes.