LStack is now in Public Beta.Anyone on a supported platform can install it.Release notes
Install · Public Beta

Three commands to your board.

Install LStack, start it, and open the dashboard. Everything after that, from models to workers to projects, you set up in the app. No account, invitation, or waitlist.

Will it run on my computer?
Yes, on Apple silicon

Macs with Apple silicon are supported. Intel Macs aren’t.

  1. 01

    Install

    Downloads the latest Public Beta over HTTPS, verifies it, and installs it in your home directory. It doesn’t start LStack.

    curl -fsSL https://lstack.ai/install.sh | bash
    View install script Opens the raw script so you can read it first.What it does
  2. 02

    Start

    Starts LStack. The control plane runs natively on your machine, not in Docker.

    lstack start
  3. 03

    Open the board

    Opens the running dashboard in your browser at its configured address. By default that’s http://127.0.0.1:4700.

    lstack open
  4. Optional

    Check readiness

    A read-only report: whether LStack is running, the address to open, where projects are stored, and whether Docker, the worker image, and on Linux the worker sandbox are ready. It changes nothing.

    lstack setup
Requirements

What you need, and when.

You can install LStack and use the dashboard straight away. Docker and a model provider come in when you’re ready to run AI workers.

  1. To install and use the dashboard
    A supported platform and a terminal.

    Create projects, manage tickets, and read local documentation without Docker.

  2. When you add AI workers
    Docker, and a model provider.

    Docker runs managed workers and container-backed project runtimes. Providers can be local, self-hosted, cloud, or a mix.

  3. For stack-mode runtimes
    Docker Compose v2 as well.

    Only projects that use stack mode need it.

Not needed
  • A global Node.js install
  • A repository checkout
  • Root or sudo to install
  • An account or invitation
Before you run it

What the installer does.

It’s a shell script you can read before you run it. These are the safeguards it applies, in order.

View install script

Opens the raw script in a new tab.

  1. 01
    Won’t run as root

    It refuses to run with root privileges and doesn’t need sudo.

  2. 02
    Checks your platform

    Unsupported platforms are refused before anything is installed.

  3. 03
    Downloads anonymously

    The release comes over HTTPS. No GitHub token, account, or other credential is required or sent.

  4. 04
    Verifies the release

    The download and its SHA-256 checksum are checked before use.

  5. 05
    Validates the archive

    Its contents are checked before anything is extracted.

  6. 06
    Installs transactionally

    Everything goes under your home directory. If installation fails, it rolls back.

After you open it

The rest happens in the app.

Installing doesn’t configure a model, worker, project, runtime, or LAN access. You set each one up from the dashboard, when you’re ready.

  1. Choose where projects are stored
  2. Sign in, or continue without an account
  3. Connect a model provider

    Local, self-hosted, or cloud.

  4. Prepare the managed worker image

    This is where Docker comes in.

  5. Create workers

    And then choose which projects may use them.

  6. Create or adopt a project
  7. Add tickets and move ready work to To Do

On Linux, if the worker sandbox needs setting up, lstack setup names the supported lstack setup-confinement command to run.

See how work moves once it’s on the board
LStack sidebar with Overview, Board, All Projects, Queue, Models / Providers, and Workers, plus a New Project button and the PackLight demo project.
Where steps 3, 5, 6, and 7 live in the sidebar
Help

If something goes wrong.

The installation guide covers most problems. If you’re still stuck, write to us.

For harder problems

You don’t need these to install. Support may ask for them if something is hard to pin down.

lstack doctor

Checks your installation and reports what it finds.

lstack doctor --bundle

Creates a redacted support bundle on your machine. It uploads nothing, excludes project data and secrets by design, and is yours to review before you decide whether to share it.