Install

works

curl -fsSL https://hobby.sh/install | bash

Piping a script into a shell asks you to trust a URL. If you would rather not, read it first or do it by hand. Every path runs the identical installer, which is deliberate and is stated at the top of the script itself. The same bytes are also mirrored at https://hobbyist.sh/install.

Requirements

Operating system	Linux or macOS. Ubuntu and Debian are the tested Linux targets. There is no Windows path
Docker	Required, not optional. Every resource is a container
Bun 1.1+	Installed for you under ~/.bun if missing. Needs no root
git	Needed to fetch the checkout
unzip	Needed by Bun’s installer. Absent from a fresh Ubuntu cloud image, and the usual reason an install fails on a cheap VPS
Memory	About 640MB of RAM and swap together, to build. See below
Filesystem	Anything works. XFS with reflinks, ZFS or APFS additionally get cheap copies. Details

On a fresh Debian or Ubuntu box, the prerequisites first:

sudo apt-get update && sudo apt-get install -y git unzip
curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker "$USER"    # then log out and back in

unzip matters more than it looks. Bun’s installer needs it to unpack the release, and Ubuntu’s cloud images do not ship it, so without it the install stops at the Bun step.

If you are on a 512MB box

Add swap before installing. The TypeScript build is the memory-hungry step, and below roughly 640MB the kernel kills it, which produces no error of its own.

sudo fallocate -l 2G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
echo '/swapfile none swap sw 0 0' | sudo tee -a /etc/fstab

Measured on one core with no swap: killed at 512MB, completed at 640MB, and at 320MB it does not fail so much as thrash for over ten minutes. 512MB with swap completes. A 1GB droplet is fine without swap.

On macOS, OrbStack is what the measurements were taken on. Docker Desktop is expected to work for everything except Caddy, whose host networking requirement is unmeasured there.

What the installer does

In order, and each step either replaces what it made last time or is a no-op, so re-running is how you upgrade:

1. Checks the box. Refuses anything that is not Linux or macOS. Requires git. Requires Docker to be installed and reachable, because an install that succeeds without it produces a daemon that cannot do the one thing it exists for.
2. Installs Bun under ~/.bun if it is missing, using Bun’s own installer. No root. A too-old Bun is rejected outright, because it gets further and then breaks somewhere less obvious.
3. Builds the workspace and the Studio bundle.
4. Installs a launcher at /usr/local/bin/hobby if that is writable, and ~/.local/bin/hobby otherwise. It is a generated shim, not a symlink, so hobby always means this build regardless of what lands on PATH later. If the directory it chose is not on your PATH, it says so.
5. Runs hobby init, which prepares ~/.hobby and checks the filesystem.

It does not install a service unit, open a firewall port, write to any system directory other than the launcher, or start a daemon. Starting the daemon is yours.

By hand

git clone https://github.com/uziiuzair/hobbyist
cd hobbyist
./install.sh

This is the same install. The one-liner’s only job is to put this checkout on disk and run this file.

Options

The bootstrap script reads three environment variables:

Variable	Default	What
HOBBY_SRC_DIR	~/.hobby/src	Where the checkout lives
HOBBY_REPO_REF	main	The branch or tag to install
HOBBY_REPO_URL	the canonical repository	Where to clone from

HOBBY_SRC_DIR=~/code/hobbyist bash <(curl -fsSL https://hobby.sh/install)

The installer itself reads HOBBY_BIN_DIR if you want the launcher somewhere specific.

Upgrading

Run the same one-liner again. An existing checkout is fetched and reset to the latest commit rather than re-cloned, which is why nothing you edit by hand under ~/.hobby/src survives an upgrade. It refuses outright if the directory is a checkout of something other than this project.

Uninstalling

There is no uninstall command, on purpose: everything it made is in two places you can see.

hobby eject <project> --release   # for each project you want to keep
rm -rf ~/.hobby                   # state, config, and every data directory
sudo rm /usr/local/bin/hobby      # or ~/.local/bin/hobby

Eject first. ~/.hobby/projects/ holds the data directories, and removing it removes your databases.

Then

Start the daemon and make something:

hobby daemon &
hobby new blog

Your first project walks through what that prints.