Skip to main content

Run OpenClaw on Raspberry Pi: Your Always-On Personal AI Assistant 

03 Feb 2026 · Originally published on x.com

I have been fascinated by the idea of having a personal AI assistant that I fully control. While cloud-based solutions are convenient, there’s something appealing about running your own AI assistant on hardware you own. With this idea, I co-authored a book - “Building a Virtual Assistant for Raspberry Pi”. I started writing the book in early 2025, and the book was published in the month of September, the that year.

AI moves fast. Really, fast. There are new models coming every other week, new AI apps, and new patterns to build AI agents.

Fast forward to January of 2026, OpenClaw (formerly Moltbot/Clawdbot), an open-source personal AI assistant, took over the internet. It promised a local-first approach to running AI assistants, with support for channels, skills, and multiple AI models. It feels closer to agentic systems!

As a tinkerer myself, I wanted to try it out, and I knew my Raspberry Pi 4 would finally get a worthy purpose.

In this tutorial, I will guide you through setting up OpenClaw on a Raspberry Pi 4. By the end, you’ll have an always-on AI assistant accessible from any device on your local network. In the next part of the blog, I will cover exposing it to the internet securely using Cloudflare Tunnels and Workers VPC.

What is OpenClaw?

OpenClaw is an open-source personal AI assistant that you run on your own devices. Unlike cloud-based assistants, OpenClaw gives you full control over your data and how your assistant operates. It supports multiple AI models (Anthropic Claude, OpenAI GPT) and can connect to various messaging channels like WhatsApp, Telegram, Discord, and more.

The key features that make OpenClaw interesting:

  • Local-first Gateway: A control plane that runs on your hardware. The AI inferecence still uses your configured provider.
  • Multi-channel support: Chat via WebChat, WhatsApp, Telegram, Slack, Discord, and more
  • Extensible: Skills and tools to extend functionality

Running OpenClaw on a Raspberry Pi makes sense for several reasons. The Pi consumes very little power (around 3-5W idle), making it perfect for an always-on setup. It’s cost-effective, and it doesn’t tie up your personal computer, so you can keep your main machine free for other tasks and prevent any potential security risks.

Prerequisites

Before you begin, ensure you have the following:

  • A Raspberry Pi 4 (I recommend the 4GB RAM model or higher)
  • A microSD card (32GB or larger recommended. My Pi is attached to an external HDD for storage. It is not required)
  • Raspberry Pi Imager installed on your computer
  • An Anthropic API key or OpenAI API key (for the AI model)

Note: OpenClaw requires Node.js 22 or higher.

Connecting to Your Raspberry Pi

If you already have Raspberry Pi OS installed and can SSH into your Pi, you can continue. But if not, follow the steps in the “Installing Raspberry Pi OS” section of the Self-Hosting n8n on a Raspberry Pi with Cloudflare Tunnels blog.

After the Pi boots up (give it a minute or two), connect to it via SSH:

Terminal window
ssh username@openclaw-pi.local

Replace username with the username you configured. If the hostname doesn’t resolve, you may need to find your Pi’s IP address from your router’s admin panel and connect using:

Terminal window
ssh username@192.168.x.x

Installing Node.js 22

OpenClaw requires Node.js version 22 or higher. The version in Raspberry Pi OS repositories is typically outdated, so we’ll install it using NodeSource.

  1. Update your system packages:
Terminal window
sudo apt update && sudo apt upgrade -y
  1. Install Node.js 22 using the NodeSource setup script:
Terminal window
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt install -y nodejs
  1. Verify the installation:
Terminal window
node --version

You should see v22.x.x or higher.

Installing OpenClaw

With Node.js installed, you can now install the OpenClaw CLI. I tried installing it globally using npm, but it didn’t work correctly. Hence, I used the following CURL command to install:

Terminal window
curl -fsSL https://openclaw.ai/install.sh | bash

Verify the installation:

Terminal window
openclaw --version

Running the Onboarding Wizard

OpenClaw comes with an onboarding wizard that walks you through the initial setup. This is the recommended way to configure your assistant.

Terminal window
openclaw onboard --install-daemon

The wizard will guide you through several configuration steps. There you can either select the Quick Start option, or select the Advanced option to customize your setup. I selected the Advanced option for more control. But for most users, Quick Start is sufficient.

1. Gateway Type

Select Local when asked about gateway type. This runs the Gateway directly on your Raspberry Pi.

Gateway is the control plane that manages your assistant. It is a WebSocket server that connects the channels, nodes, sessions, etc.

If you already have a Gateway running elsewhere, you can connect to it by selecting Remote type. Since you are running everything on the Pi, you select Local.

2. Authentication

You’ll need to configure how OpenClaw authenticates with AI providers. For a headless Raspberry Pi setup, I recommend using an API key rather than OAuth (which requires a browser).

When prompted, select your preferred provider:

  • Anthropic (recommended): Paste your Anthropic API key
  • OpenAI: Paste your OpenAI API key

I have been trying out the MiniMax M2.1 model, and it works great. Hence, I selected minimax.io as my provider and pasted the API key.

Note: Keep your API keys secure. Never share them publicly or commit them to version control.

3. Channels

For this tutorial, we’ll use the built-in WebChat. You can skip channel configuration for now and add WhatsApp, Telegram, or other channels later.

4. Daemon Installation

When asked about installing the daemon, select Yes. This installs OpenClaw as a systemd service that starts automatically when your Pi boots.

Verifying the Installation

After the onboarding wizard completes, verify that everything is working:

  1. Check the Gateway status:
Terminal window
openclaw gateway status

You should see that the Gateway is running.

  1. Run the health check:
Terminal window
openclaw health

This checks connectivity to your configured AI provider.

  1. Run the doctor command to check for any issues:
Terminal window
openclaw doctor

If the doctor reports any issues, follow the suggestions to resolve them.

Accessing the Control UI

The OpenClaw Gateway serves a Control UI interface that you can access from any device on your local network.

  1. Find your Raspberry Pi’s IP address:
Terminal window
hostname -I

Note the first IP address (e.g., 192.168.1.100).

  1. Open a browser on any device connected to your local network and navigate to:
http://192.168.1.100:18789

Or if mDNS is working on your network:

http://openclaw-pi.local:18789

You should see the OpenClaw Control UI. If you see an error, you might need to enable the Control UI in your configuration file.

Authenticating with the Gateway

If you configured a Gateway token during onboarding (recommended for security), you’ll need to enter it:

  1. Click on Overview in the Control UI
  2. Paste your Gateway token
  3. The connection status should change to “Connected”

If you don’t remember your token, you can find it in your configuration:

Terminal window
cat ~/.openclaw/openclaw.json | grep token

Starting a Conversation

Once connected, you can start chatting with your AI assistant directly in the browser. Type a message and press Enter. The assistant will respond using the AI model you configured.

Managing the OpenClaw Service

Since we installed OpenClaw as a systemd service, it will start automatically when your Raspberry Pi boots. Here are some useful commands for managing the service:

Check the service status:

Terminal window
systemctl --user status openclaw-gateway

Restart the service:

Terminal window
systemctl --user restart openclaw-gateway

Stop the service:

Terminal window
systemctl --user stop openclaw-gateway

Configuration Tips for Raspberry Pi

Here are some configuration tweaks I found helpful when running OpenClaw on a Raspberry Pi:

Binding to All Interfaces

By default, OpenClaw may bind only to localhost. To access it from other devices on your network, ensure your configuration allows LAN access. Check your ~/.openclaw/openclaw.json:

{
  "gateway": {
    "bind": "lan"
  }
}

Note: When binding to LAN, ensure you have proper authentication (Gateway token) enabled for security.

After changing the configuration, restart the service:

Terminal window
systemctl --user restart openclaw-gateway

What’s Next?

You now have a fully functional personal AI assistant running on your Raspberry Pi. Here are some ideas for extending your setup:

  • Add messaging channels: Configure WhatsApp or Telegram to chat with your assistant from your phone. See the OpenClaw channels documentation.
  • Explore skills: OpenClaw supports skills that extend its capabilities. Check out ClawHub for available skills.

Conclusion

You have successfully set up OpenClaw on your Raspberry Pi as an always-on personal AI assistant. The setup is local-first, privacy-respecting, and accessible from any device on your network. The best part? Your Raspberry Pi sips power while providing you with a capable AI assistant 24/7. In the next part of this blog, I will cover how to expose your OpenClaw assistant to the internet securely using Cloudflare Tunnels and Workers VPC.

I’m excited to continue exploring what’s possible with this setup. If you run into any issues or have questions, feel free to reach out on LinkedIn or X (Twitter).

References

Last updated on 03 Feb 2026