OpenClaw on Raspberry Pi: Your $35 Always-On AI Assistant

Why Raspberry Pi for OpenClaw?

OpenClaw is designed to run wherever Node.js 22 (or 20+) runs. On a Raspberry Pi 4 with 4GB RAM or a Pi 5 with 8GB, you have enough memory for the Node process plus the gateway and any skills you enable. The key is to use cloud API mode: your Pi sends prompts to Claude or OpenAI and receives replies; it does not need to run large local models. That keeps latency low and hardware requirements minimal. Users report success on Pi 4 (4GB), and the Pi 5 with 8GB is officially tested and recommended by the community for a smoother experience.

Benefits of a Pi-based OpenClaw include low power consumption (typically under 10W), silent operation, and the ability to leave it in a corner or on a shelf. You get a dedicated always-on agent without tying up your main computer. A fast SD card (A2) or USB SSD boot is recommended to avoid I/O bottlenecks, and a heatsink or small fan helps under sustained load.

1 Prepare Raspberry Pi OS

Use the official Raspberry Pi Imager to flash 64-bit Raspberry Pi OS. Choose the Lite image if you are comfortable with the command line; it uses less RAM than the desktop version and leaves more headroom for OpenClaw. After first boot, run sudo raspi-config to set locale, timezone, and optionally enable SSH. Update the system: sudo apt update && sudo apt upgrade -y.

2 Install Node.js 20+ (or 22)

OpenClaw requires Node.js 20 or newer; Node 22 is preferred. On Raspberry Pi OS, the default apt packages may be older. Use the NodeSource repository for ARM64:

Confirm the version is v20.x or higher. If Node 22 binaries are listed for your architecture, use setup_22.x instead for the best compatibility with the latest OpenClaw releases.

3 Install OpenClaw

The official one-line installer works on Raspberry Pi OS (Debian-based):

Alternatively, install globally with npm: sudo npm install -g openclaw@latest. Then run openclaw --version to confirm. Next, start the onboarding wizard so OpenClaw can connect to your AI provider and your first messaging channel (e.g. Telegram):

Follow the prompts to choose your model (Claude, GPT-4o, or Gemini), add your API key, and create or link a Telegram bot. The --install-daemon flag installs a systemd user service so OpenClaw starts on boot and restarts if it crashes.

Low-Memory and Performance Tips

On a 4GB Pi, limit the number of skills you enable and avoid memory-heavy options. You can tune the OpenClaw config (under ~/.openclaw/) to reduce concurrency and cache size if needed. Keep conversation history to a reasonable limit (e.g. 100 messages) to avoid unbounded growth. For AI inference, stick to cloud APIs; running Ollama with large models on a Pi is possible on Pi 5 8GB but will be slow for anything beyond small models (e.g. 3B parameters).

If you use Telegram, Discord, or WhatsApp, traffic goes over the internet. In regions where these services or your chosen AI API are throttled or unreliable, a stable VPN on the Pi (or on your router) can keep your OpenClaw responsive. Choose a VPN that supports Linux and can run as a systemd service so it stays up alongside OpenClaw.

Security and Updates

OpenClaw has access to your shell and can run skills that interact with the system. Review the official security documentation before exposing the gateway. Keep the OS and OpenClaw updated: sudo apt update && sudo apt upgrade -y and sudo npm update -g openclaw, then restart the daemon.

FAQ