How to Install OpenClaw and Hermes Agent on Qualcomm Arduino Boards, Rubik Pi 3 and Snapdragon PCs

This blog post was originally published at Qualcomm’s website. It is reprinted here with the permission of Qualcomm.

 

A Step-by-Step Guide with Arduino UNO Q, Rubik Pi 3, and PCs with Snapdragon

Introduction

Welcome, developers and tech enthusiasts! This blog post will guide you through running OpenClaw and Hermes Agent on Qualcomm Technologies’ platforms, using Arduino UNO Q, Rubik Pi 3 and a PC with Snapdragon.

Here’s what you’ll learn in this walkthrough, with hands-on examples:

• What OpenClaw and Hermes Agent are

• How to set up dependencies

• Installation steps

• Large language model (LLM) integrations

Whether you’re building smart robotics or experimenting with AI at the edge, this guide is designed to help you get started efficiently.

OpenClaw and Hermes Agent: What is this?

OpenClaw and Hermes Agent are two of the most talked-about open-source projects right now. They’re autonomous, personal agents that can complete tasks, log into websites, and create and edit files—much like Claude Cowork, but able to work with any cloud AI model provider, not just Claude.

These open-source frameworks are designed to bridge hardware platforms and AI models, making it easy to deploy intelligent applications on devices like microcontrollers and single-board computers. They enable seamless communication between edge devices and LLMs.

And both of them run well on a range of Qualcomm Technologies and Snapdragon platforms, from sub-$100 single-board computers to high-end laptops with Snapdragon X2 Elite Extreme.

OpenClaw, Hermes and their like are orchestrators. They don’t run an AI model themselves; they are negotiators between AI models and everything else on your system. They are relatively lightweight and can run smoothly in the background with whatever else you’re doing.

What can I do with OpenClaw?

There’s a lot to be said for using AI agents to simplify tasks like doing research, creating reports, summarizing long documents and planning travel. But the tokens add up fast, and the usage limits get frustrating. And, of course, there’s the security angle: You don’t want an agent running around in your personal and business data like a bull in a china shop.

Enter the Arduino UNO Q and RUBIK Pi 3. They are lightweight IoT devices, much less expensive than your personal or business laptop but powerful enough to run desktop Linux. You can configure them with limited credentials and they’ll still surprise you with how much time and work they save you. Plus, by running them in a hardware sandbox – the best kind – you limit the potential for unintended consequences because they keep AI agents away from the rest of your data.

OpenClaw Dependencies

OpenClaw depends on the following software:

• Core Infrastructure: Docker is the primary dependency for containerized deployment, ensuring a stable environment.
• Runtime Environment: Node.js (v22 or higher) is essential, with Node Version Manager (nvm) recommended for management.
• Scripting/Automation:  Python 3.8+ is needed for core logic, while libraries such as PyYAML, Requests, and GitPython are required for configuration and Git operations.
• Access to AI Models: OpenClaw supports popular LLMs from families such as Anthropic, OpenAI GPT, Google Gemini and xAI Grok, among others.

NODE acts as the middleware, managing device communication and workflow orchestration. The LLM powers intelligent responses and controls. The AI model can be hosted in the cloud or locally.

Arduino UNO Q: Introduction

What is Arduino UNO Q

The Arduino UNO Q is a single-board computer, a dual-brain board built around the Qualcomm QRB2210 SoC, offering a robust platform for embedded development. Its integrated Wi‑Fi and efficient power management make it suitable for edge‑connected prototypes, and it is affordable and easily accessible.

It comes in a version with 4GB of memory and 32GB of storage, which are adequate specs for running OpenClaw or Hermes Agent. Because the board runs a full Debian OS, you can easily install and run frameworks like OpenClaw and Hermes, take advantage of standard Linux tooling, and deploy software using familiar package management workflows.

Benefits of installing your agent on Arduino UNO Q

Installing your agent on a low-cost, discrete board has advantages. Most importantly, the agent does not have access to your main machine or the data on it; it has access to only the accounts and data you give it. (You can, of course, give it SSH access to your other machines, at your own risk.)

Arduino UNO Q is easy to set up thanks to the Arduino App Lab software, which configures the board and lets you use your PC as a terminal.

Arduino UNO Q system partitioning

There’s one idiosyncrasy about installing these agents on the Arduino UNO Q. Most of the board’s free storage is in a user partition under /home/Arduino, shipping with separate system (root) and user partitions. So it’s important to install dependencies such as npm, homebrew, and uv under /home/Arduino, which lives on the user partition, rather than in /bin, /usr or even /home.

The system partition is intentionally small and can’t hold the full set of Homebrew packages or OpenClaw skill dependencies. To avoid running out of space, everything should be installed under /home/Arduino.

The instructions below help guide you to put the dependencies in the right places. With correct installation, you will still have more than 10GB available for skills and data.

After installing OpenClaw and its skills, you may see warnings about missing npm packages for certain connectors. That’s expected—just install the requested packages manually and continue.

The installation process on other IoT boards with processors from Qualcomm—such as the Thundercomm Rubik Pi 3, Radxa Q6A, and the upcoming Arduino VENTUNO Q—should be even easier. They have more RAM and storage, so you won’t need to worry about where the dependencies are installed.

If you’re feeling creative, you can also explore this how-to from David Groom, who paired OpenClaw with the latest 4GB Arduino UNO Q to make embedded hardware accessible conversationally, without having to write any code.

Arduino UNO Q: Installing OpenClaw

Below is a clean, reproducible setup workflow.

1. Create a local directory for user‑space tools

mkdir -p /home/arduino/.local

2. Install Node.js (ARM64)

cd /tmp
curl -fsSL https://nodejs.org/dist/v22.22.2/node-v22.22.2-linux-arm64.tar.xz -o node.tar.xz

Extract Node.js into your user‑local directory:

tar -xJf node.tar.xz -C /home/arduino/.local --strip-components=1

Add Node.js to your PATH:

echo 'export PATH=/home/arduino/.local/bin:$PATH' >> ~/.bashrc
source ~/.bashrc

Verify:

node --version
npm --version

Configure npm to install packages in your home directory:

npm config set prefix /home/arduino/.npm-global
npm config set cache /home/arduino/.npm
echo 'export PATH=/home/arduino/.npm-global/bin:$PATH' >> ~/.bashrc
source ~/.bashrc

3. Configure Homebrew to install under /home/arduino

Set Homebrew’s cache location:

echo 'export HOMEBREW_CACHE=/home/arduino/.cache/Homebrew' >> ~/.bashrc

Export Homebrew paths for this session:

export HOMEBREW_PREFIX=/home/arduino/.linuxbrew
export HOMEBREW_CACHE=/home/arduino/.cache/Homebrew
export HOMEBREW_LOGS=/home/arduino/.cache/Homebrew/Logs

Install Homebrew into your user partition:

git clone --depth=1 https://github.com/Homebrew/brew.git /home/arduino/.linuxbrew/Homebrew
mkdir -p /home/arduino/.linuxbrew/bin
ln -s /home/arduino/.linuxbrew/Homebrew/bin/brew /home/arduino/.linuxbrew/bin/brew

Add Homebrew to PATH:

echo 'export PATH=/home/arduino/.linuxbrew/bin:$PATH' >> ~/.bashrc
source ~/.bashrc

Run initial setup:

brew update

Verify:

brew --prefix

4. Install and configure OpenClaw

npm install -g openclaw@latest
openclaw config

See below for OpenClaw installation screenshots. Once your UNO Q is ready, you can connect it to NODE and start building AI-powered projects.

Arduino UNO Q: Installing Hermes Agent

## Prepare Redirected Home Directories

# Redirect everything large to /home/arduino

export HERMES_HOME="/home/arduino/.hermes"
export UV_HOME="/home/arduino/.local"
export UV_CACHE_DIR="/home/arduino/.local/cache/uv"
export UV_TOOL_DIR="/home/arduino/.local/bin"

# Create the directories

mkdir -p /home/arduino/.local/bin
mkdir -p /home/arduino/.local/share/uv
mkdir -p /home/arduino/.local/cache/uv
mkdir -p /home/arduino/.local/python
mkdir -p /home/arduino/.hermes 

cat >> ~/.bashrc << 'EOF'

# Hermes Agent on Arduino Uno Q — redirect large dirs to /home/arduino

export HERMES_HOME="/home/arduino/.hermes"
export UV_HOME="/home/arduino/.local"
export UV_CACHE_DIR="/home/arduino/.local/cache/uv"
export UV_TOOL_DIR="/home/arduino/.local/bin"
export UV_PYTHON_INSTALL_DIR="/home/arduino/.local/python"

# Ensure ~/.local/bin is on PATH

export PATH="/home/arduino/.local/bin:$PATH"
EOF
source ~/.bashrc

## Install Node.js Under `/home/arduino`

# Download the ARM64 Node.js v22 LTS tarball

cd /home/arduino
curl -fsSL https://nodejs.org/dist/v22.22.2/node-v22.22.2-linux-arm64.tar.xz -o node.tar.xz

# Extract in place

tar xJf node.tar.xz

# Move to final location

mv node-v22.22.2-linux-arm64 /home/arduino/node

# Symlink binaries to ~/.local/bin (which is on PATH)

ln -sf /home/arduino/node/bin/node /home/arduino/.local/bin/node
ln -sf /home/arduino/node/bin/npm /home/arduino/.local/bin/npm
ln -sf /home/arduino/node/bin/npx /home/arduino/.local/bin/npx

# Clean up

rm node.tar.xz

## Install Hermes Agent with Redirected Paths

export HERMES_HOME="/home/arduino/.hermes"
export UV_HOME="/home/arduino/.local"
export UV_CACHE_DIR="/home/arduino/.local/cache/uv"
export UV_TOOL_DIR="/home/arduino/.local/bin"
export UV_PYTHON_INSTALL_DIR="/home/arduino/.local/python"
export PATH="/home/arduino/.local/bin:$PATH"

# Run the installer, directing everything to /home/arduino

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash -s -- \
  --dir /home/arduino/hermes-agent \
  --skip-setup

## Configure Hermes

hermes setup

## Start Chatting

hermes

As one of your first queries, ask:

Can you make sure to install skills and their binaries under /home/arduino and not in root directories? Symlink from the root directories if needed. Remember this and do it going forward. Also, instead of /tmp/, use /home/arduino/tmp/ – remember this across all sessions going forward.

Rubik Pi 3: Introduction

If you plan to run OpenClaw in more-demanding scenarios, consider the RUBIK Pi 3, an embedded compute module built by Thundercomm around the Qualcomm® Dragonwing QCS6490 SoC.

The RUBIK Pi 3 is a high‑performance platform with multi-core CPU/GPU processing, integrated AI acceleration, modern connectivity options and ample I/O options. It runs a full Ubuntu distribution from Canonical, offering a familiar Linux environment with access to standard tooling, package management, and container workflows. That makes the RUBIK Pi 3 a solid choice for edge AI prototyping, robotics and any project that benefits from a production‑grade Linux stack on compact hardware.

Rubik Pi 3: Installing OpenClaw

1. Prepare the Rubik Pi 3

1. Download the latest Rubik Pi 3 OS image from the official website.
   Install Ubuntu on Qualcomm IoT Platforms | Ubuntu
2. Follow instructions to install Ubuntu 24.04 for Rubik Pi 3
3. Complete the initial OS setup, connect to Wi-Fi.
4. Update the package repositories with

sudo apt update && sudo apt upgrade

5. Install Git

sudo apt update && sudo apt install -y git

6. You can see more information and getting started instructions for RUBIK Pi 3 Development Board.

2. Install Node.js

It’s recommended to use nvm (Node Version Manager) to install and manage Node.js versions.

# If curl is not available
sudo apt update 
sudo apt install curl -y
# Download and install nvm
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.3/install.sh | bash
# Load nvm into the current terminal (no need to restart terminal)
\. "$HOME/.nvm/nvm.sh"
# Install Node.js (v22 or later)
nvm install 22
# Verify installation
node -v   # Display v22.x.x or higher
npm -v    # Display 10.x.x or higher

3. Install OpenClaw

There are two ways to install OpenClaw.

1.      One-click installation script (recommended)

curl -fsSL https://openclaw.bot/install.sh | bash

The script will automatically download and configure OpenClaw. Installation typically takes 5–10 minutes, depending on network speed. If it’s not complete after 15 minutes, consider switching to option 2.

2.      Install via npm

npm i -g openclaw

Verify installation:

openclaw --version

Installation process screenshots:

Figure 1: Screenshot – OpenClaw Installer

 

If you see the OpenClaw onboarding information shown below, then you have successfully installed OpenClaw. You may proceed with configuration.

Figure 2: Screenshots: OpenClaw onboarding screens

Figure 3: Screenshot: OpenClaw model/auth provider

If you are deploying OpenClaw locally, ensure the LLM server is running on a laptop powered by Snapdragon X Elite. Use API keys or authentication tokens to secure communication.

Here is more documentation showing how to run OpenClaw on the RUBIK Pi 3 using the MiniMax M2.1 model from MiniMax Open Platform. Once configured, OpenClaw can send queries to the LLM and receive intelligent responses, powering automation or smart interactions.

Installing OpenClaw and Hermes Agent on PCs with Snapdragon X

On PCs with X Series processors—whether Snapdragon X, Snapdragon X Plus, Snapdragon X Elite, or Snapdragon X2—AI orchestrators run smoothly. If you want a polished Windows experience with a great-looking interface, choose Claude Cowork, which runs natively on PCs with Snapdragon.

To install OpenClaw or Hermes Agent on your PC with Snapdragon, use Windows Subsystem for Linux. WSL is a full Linux environment available for every PC with Snapdragon. You can install it by opening the Powershell app on your PC and typing wsl –install. The system defaults to Ubuntu 24.04 LTS, which is fine for these applications.

Once it’s installed, type wsl in Powershell to enter the Linux environment and set up your agent according to each agent’s official guide. No special instructions apply.

What’s next?

Running OpenClaw on Qualcomm Technologies platforms unlocks new opportunities for AI-driven hardware projects. With clear installation steps for Arduino UNO Q and Rubik Pi 3, straightforward dependency setup, and seamless LLM integration, you can quickly build and deploy intelligent applications.

Remember to start small, test often, and keep exploring.

You should be aware that OpenClaw and Hermes Agent will have access to your local files! As a result, many online sources suggest that, at a minimum, you take the following precautions:

• Use Docker to insulate the agent from your primary filesystem.
• Store API keys as environment variables rather than as plain text in the config file.
• Place strict spending limits on any API key to which your agent has access.
• Do not accept external connections to your gateway.
• Make sure messaging channels accept messages only from you specifically.
• Monitor logs for any untoward action.

Enjoy your agentic experience and happy coding!

Authors:
Sascha Segan, Sr. Manager, Public Relations, Qualcomm Technologies, Inc.

Rajan Mistry, Engineer, Senior Staff, Qualcomm Technologies, Inc.