Skip to main content

​
Requirements

  • Linux
  • Python 3.13+ (the quick install below provides it automatically)
  • (Optional) NVIDIA Jetson — if ZED cameras are used.
One command installs everything and starts the web control panel: 
curl https://axol.almond.bot/install -fsS | bash
This:
  • Installs uv if it isn’t already installed.
  • Installs the axol CLI from GitHub as a uv tool (Python 3.13 included), with every dependency extra except cuda and dev.
  • Registers a root systemd service (axol.service) that runs axol serve at startup and restarts it automatically if it ever goes down.
When it finishes, open axol.almond.bot and connect to the machine — no terminal needed from then on. The install keeps itself in sync with the latest release: whenever the control panel connects, the server checks for updates in the background and restarts onto the new version once idle. (Re-running the install command also upgrades.) 
journalctl -u axol -f        # service logs
systemctl restart axol       # manual restart
The service runs as root, so privileged steps (CAN bring-up, PTP clock-sync daemons) need no sudo password.

​
Development install

For working on Axol itself, install from a clone using uv. pyroki and lerobot are sourced from Git and resolved automatically: 
uv sync
Then activate the virtual environment so the axol CLI is on your path: 
source .venv/bin/activate
You need to activate the environment in every new shell session. Alternatively, prefix every command with uv run (e.g. uv run axol can.setup) to skip activation.

​
Optional extras

Install optional dependency groups as needed (the quick install above includes all of them except cuda and dev):
ExtraContentsWhen to use
lerobotLeRobot (from GitHub)collect-data, run-policy
simviserteleop --sim
videoaiortcStreaming the ZED camera feeds to the headset over WebRTC (teleop --zed_host, collect-data) — see Camera views
cudaJAX with CUDA 13 supportGPU-accelerated JAX (IK solver used by teleop); note that CPU is usually faster for the JAX IK solver
devOpenCV (headless)Development / debugging
uv sync --extra lerobot --extra sim --extra video   # teleoperation + data collection
uv sync --extra lerobot --extra cuda                # policy execution on GPU
uv sync --extra lerobot --extra sim --extra video --extra cuda   # everything

​
ZED Python bindings

The ZED Python bindings (pyzed) are not on PyPI and must be installed separately after the ZED SDK is installed: 
axol zed.install
See zed.install for details.

​
Web control panel (optional)

With the quick install, the control panel is already running — just open axol.almond.bot and connect to the machine. For a development install, you can build and serve the UI locally instead. This produces the bundle that axol serve hosts at web/app/dist: 
cd web
npm install
npm run build --workspace=packages/axol-vr-client   # client package first
npm run build --workspace=app                        # → web/app/dist
Then start the server and open the printed URL: 
axol serve
The same build also serves the VR interface. axol serve works without this step (the API is still available) — the hosted UI at axol.almond.bot connects to it directly. See the Web Control Panel guide.

​
First-run hardware setup

Before using any motor or robot commands, initialize the CAN hardware: 
axol can.setup
Run directly in a terminal, sudo is invoked automatically where required (with the quick install, the control panel runs this for you as root on first connect). See can.setup for what it configures.