Run OpenClaw on a Raspberry Pi
A Raspberry Pi works as a quiet, always-on OpenClaw host, as long as the models run somewhere else.
This guide assumes the models run through a cloud API. Even the smallest local llama.cpp recipe needs about 8 GiB of memory.
Pick the board and the OS
The docs rate the Pi 5 best and a Pi 4 with 4 GB as the sweet spot. You need a 64-bit OS and at least 1 GB RAM; 2 GB or more is recommended, together with a USB SSD.
Run the installer
This is the same script as on any Linux host. It provisions Node 24 LTS and the openclaw CLI, then starts onboarding. Confirm that the Pi is running the 64-bit OS before you start.
shell curl -fsSL https://openclaw.ai/install.sh | bashOnboard and install the service
If you skipped the wizard, run it now. It writes the config and the gateway token, and --install-daemon sets up the systemd user service so the gateway runs in the background.
shell openclaw onboard --install-daemonKeep it private and light
Leave gateway.bind on loopback, even on a home network, and reach the dashboard through an SSH tunnel or Tailscale. Point the agent at a cloud provider, or at a model server on a more powerful machine, instead of running models on the Pi. Store OpenClaw on a USB SSD, as the docs recommend.