Quickstart

Choose how to run Swarm.

Local

Local.

01

Install Swarm

Linux x86_64 and WSL Ubuntu are available now through the latest-release installer. You do not need to clone the source repository to install Swarm.

curl -fsSL https://swarmagent.dev/install | sh
02

Start Swarm

Run the terminal UI from your current directory, or open Swarm Desktop in the browser.

swarm Starts from your current directory.
swarm --desktop Opens Swarm Desktop in the browser, using a workspace-first launcher.
03

Connect a provider

Run auth to connect an AI provider and start a session.

/auth

Swarm can run in multiple network configurations locally.

Configuration

Container

Container.

01

Install Podman or Docker

Local child swarms use Podman or Docker. Install one of those runtimes before using container-based Add Swarm workflows.

02

Launch managed local child swarms

From an existing local Swarm, Add Swarm can launch a child swarm into Docker or Podman, replicate selected workspaces, and keep the child managed by the parent.

03

Make local container setups network reachable

Local Swarm and local child containers can also be configured for LAN or VPN-reachable setups. Use swarm mode settings to publish a reachable callback endpoint instead of loopback-only defaults.

See the configuration reference for local child transport, startup mode, and advertised callback settings.

Remote

Remote.

01

Use a remote Linux machine over SSH

The remote path prepares the image and payload locally, sends the selected workspaces and config over SSH, then attaches the remote child back to the current Swarm.

02

Choose the callback transport

Remote child deploy supports callback over a reachable endpoint. SSH installs the child; after launch, the child still needs a real callback path back to the master through the configured advertise_host/advertise_port.

03

Treat Remote as a host workflow

Remote means a Linux host you control, typically a VM or VPS. It is not a separate provider-specific runtime.