QemuClaw

One-click isolated deployment for OpenClaw — a personal AI assistant running securely inside a QEMU virtual machine.

Think of it as an aquaculture facility for your OpenClaw: a contained environment that keeps your OpenClaw thriving without damaging your furniture (main system). With 1,000+ known vulnerabilities — including 4+ critical ones enabling potential remote code execution — isolation is essential.

Download

Desktop App

Platform	Download
Windows	QemuClaw-Setup-1.0.4.exe (includes QEMU + 7-Zip)
macOS	QemuClaw-1.0.4.dmg
Linux	QemuClaw-1.0.4-linux.AppImage
Linux (deb)	QemuClaw-1.0.4-linux.deb

VM Image

Download OpenClaw VM Image (vm-headless-25477445209)

Install (Desktop App)

Windows: Download and run the .exe installer. QEMU and 7-Zip are bundled — no extra setup.

macOS: Install QEMU first (brew install qemu), then open the .dmg.

Linux: Install QEMU first (sudo apt install qemu-system-x86), then run the .AppImage or install the .deb.

On first launch, the app will:

1. Ask you to locate QEMU if it's not bundled or found automatically
2. Show a setup wizard for VM resources (memory, CPU, shared folder)
3. Download the VM image if not already present
4. Boot the VM headless — a status window shows boot progress
5. Open the terminal for OpenClaw onboarding setup
6. After setup, access OpenClaw via the Web UI (system tray > Open Web UI)

The app runs in the system tray. Right-click the tray icon for all options.

The icon in system tray

Manual Quick Start (without Desktop App)

If you prefer to run QEMU directly:

1. Install QEMU
2. Download the VM image from the releases page (.qcow2 file, or .tar.gz.* split files)
3. If split files: cat *.tar.gz.* | tar xvfz -
4. Start the VM:

   qemu-system-x86_64 -m 2G -smp 2 \
     -drive file=./openclaw-headless-compressed.qcow2,format=qcow2

5. Login with username node and password openclaw
6. Run cd /app && node dist/index.js onboard

Development

git clone https://github.com/NewJerseyStyle/QemuClaw.git
cd QemuClaw
npm install
npm start

Requirements for development:

• Node.js 18+
• QEMU installed and on PATH (or set path in the app on first launch)
• 7-Zip installed (for VM image extraction on Windows)

Using Local LLMs (Ollama, LM Studio, etc.)

If you run a local LLM provider on your host machine, the VM can access it via the hostname host.local (resolves to 10.0.2.2, the QEMU user-mode networking gateway to the host).

Host service	URL from inside the VM
Ollama (localhost:11434)	http://host.local:11434
LM Studio (localhost:1234)	http://host.local:1234
Any host service on port N	http://host.local:N

During OpenClaw onboarding, when asked for an API endpoint, use http://host.local:<port> instead of http://localhost:<port>.

Architecture

QemuClaw (Electron app)
  |
  |-- QEMU (headless, -display none)
  |     |-- Serial console (TCP) --> xterm.js terminal in app
  |     |-- QMP control (TCP)    --> graceful shutdown, status
  |     |-- Port forwarding      --> localhost:18789 (OpenClaw Web UI)
  |     \-- Shared folder        --> SMB (Windows) / 9p (Linux/macOS)
  |
  \-- System tray
        |-- Open Web UI
        |-- Open Terminal
        |-- Update OpenClaw
        |-- Restart VM
        |-- Settings
        \-- Quit

License

MIT

Acknowledgements

• 🙏 Thank you to Arthur for valuable support and assistance.
• 🙌 Thanks Hong Kong School of AI ⓕ for getting this project off to a strong start.
• 💖 Grateful to OpenClaw for the impressive personal AI assistant.
• 🌟 Thanks to the Alpine Community for providing a handy Docker image.
• 🤗 Appreciation to QEMU for the amazing virtualization technology.
• 👉 A nod to 7-zip for a powerful utility.