🦞 OpenClaw VM

A cross platform virtual machine for running OpenClaw, the open-source AI agent. Provides a hardened sandbox with firewall rules, optional local inference via Ollama, and remote access over SSH.

Features

Quick Setup

• One-Command Install — Get up and running instantly. A simple curl script automates the installation of VirtualBox, Vagrant, and the CLI.
• Cross-Platform — Native support for macOS (Intel & Apple Silicon) and Linux (Debian, Fedora, Arch) with automatic architecture detection.

Secure

• Isolation — A full VM boundary with no shared folders and a hardened UFW firewall keeps your host system untouchable.
• Sandboxed Secrets — Agent skills are strictly restricted; they cannot access host files or internal VM credentials.

Modular

• Modular Inference — Install Ollama only if you want local models. Skip it to save disk space and RAM when leveraging cloud providers.
• Hot-Swapping — Change models, add skills, or update API keys from inside the VM without re-provisioning.

Remote Access

• Headless Hosting — Deploy on a Mac Mini or home server and SSH in from any device on your local network.

---

How to Run

1. Install

curl -sSL https://raw.githubusercontent.com/DaevMithran/openclaw-vm/main/install.sh | bash

This detects your OS and installs VirtualBox and Vagrant using the appropriate package manager (Homebrew on macOS, apt/dnf/pacman on Linux), then clones the repo and adds openclaw-vm to your PATH.

2. Start the VM

openclaw-vm start

3. Configure OpenClaw

openclaw-vm ssh
openclaw onboard          # Interactive setup (first time)

Or skip onboard by providing an existing config:

openclaw-vm start --config ~/openclaw.json

Validate your config at any time:

openclaw config validate

4. Start the Gateway

openclaw-vm ssh
openclaw gateway          # Start the gateway
openclaw status           # Check it's running

5. Interact

Visit http://localhost:18789 in your browser for WebChat, or use openclaw tui inside the VM for a terminal interface.

Config Flags

Use flags with openclaw-vm config for VM resources:

openclaw-vm config --ram 16384 --cpus 8
openclaw-vm config --ollama
openclaw-vm config --no-ollama
openclaw-vm config --password my-secure-password

# Apply changes to a running VM
openclaw-vm provision

OpenClaw Configuration

Once inside the VM, use OpenClaw's built-in commands to configure models, API keys, skills, and channels:

openclaw onboard                   # Interactive setup (first time)
openclaw config                    # Validate config
openclaw gateway                   # Start the gateway
openclaw status                    # Check status
openclaw tui                       # Terminal chat interface

Remote Access

Host the VM on a Mac Mini or server and SSH in from other devices on your LAN.

Setup

Set a SSH password via the interactive wizard or CLI flags:

openclaw-vm config --password your-strong-password
openclaw-vm provision

Then connect from any machine on the network:

ssh vagrant@<host-ip> -p 2222

Password auth is opt-in. If no password is set, SSH uses Vagrant key-only (host only). When enabled, SSH is hardened automatically: root login disabled, max 3 auth attempts.

Headless Mode

The VM runs headlessly by default (no GUI window). Deploy it on a Mac Mini or home server and manage it entirely over SSH:

# On the server
openclaw-vm start
openclaw-vm ssh -c "openclaw gateway"

Project Structure

openclaw-vm/
├── openclaw-vm              # CLI wrapper (start, stop, ssh, config, provision)
├── install.sh               # One-line installer (curl | bash)
├── Vagrantfile              # VM definition (VirtualBox / ARM64)
├── config/                  # Generated at first run (gitignored)
├── scripts/
│   ├── provision.sh         # System provisioning (root): Node, firewall, SSH, optional Ollama
│   └── user-setup.sh        # User setup: OpenClaw install, workspace
└── README.md

Star History

Contributing

Contributions are welcome. Please open an issue first to discuss what you'd like to change.

License

MIT

Support This Project

Like this project? Donate via PayPal and support continued improvements.