Installing the OpenClaw AI agent runtime has one non-negotiable technical prerequisite: You must have Node.js version 22 or higher installed to support its core engine. This isn’t a suggestion—it’s a hard requirement for its advanced features like persistent memory, multi-agent personas, and extensible skills to function correctly. This guide breaks down not just the “why,” but the exact steps to get OpenClaw (formerly Moltbot/Clawdbot) running securely, whether on your local machine or a 24/7 VPS.
Node.js 22+ is Non-Negotiable for OpenClaw’s Core Engine
The entire OpenClaw system is built on modern Node.js APIs and performance features exclusive to version 22 and above. Attempting to install it on Node 18 or 20 will result in immediate failure. Always verify your version first.
- Command to Check: Run
node -vin your terminal. - Mandatory Result: The output must show v22.x.x or higher (e.g., v22.11.0).
- Why It Matters: Version 22+ ensures compatibility with critical libraries and provides the runtime stability needed for automated email triage, smart home control, and complex workflow agents.
Think of Node.js 22 as the foundational operating system for OpenClaw. Without it, the advanced architecture simply cannot initialize or run.
Installation Method 1: Local Linux CLI in 4 Straightforward Steps
For development or testing, installing OpenClaw directly on a Linux machine is the fastest path. This method gives you full control but requires careful security handling.
Step 1: Install Node.js 22 or Higher
If node -v shows an older version, you must upgrade. Use your system’s package manager (like apt for Ubuntu) or download directly from the official Node.js website.
Step 2: Install OpenClaw Globally via npm
With the correct Node version, install the gateway globally using the command: npm install -g openclaw. You can also use pnpm or bun.
Step 3: Launch the Setup Wizard & Install the Daemon
Run openclaw onboard --install-daemon. This crucial command starts the configuration wizard and installs a systemd service to run OpenClaw in the background 24/7.
Step 4: Configure Your Gateway During the Wizard
- Gateway Type: Choose “local gateway”.
- Workspace: Defaults to
~/.openclaw. - Model Provider: Select Anthropic (Claude), OpenAI, or Google and paste your API key.
- Gateway Bind: For local use only, select “loopback”. For remote access, choose “tail” (Tailscale).
- Gateway Token: Leave blank to auto-generate a secure token.
The local CLI install is ideal for technical users who need direct file system access and are comfortable with terminal-based security management.
Installation Method 2: VPS with Docker (Recommended for 24/7 Uptime)
For a permanent, secure, and remotely accessible installation, deploying on a Virtual Private Server (VPS) using Docker is the professional standard.
VPS Setup: Use Hostinger KVM2 with Linux LTS
Provision a VPS (like Hostinger’s KVM2 plan) with a long-term support (LTS) Linux OS. This gives you a clean, isolated environment.
Deploy via Docker Catalog in 3 Clicks
- In your VPS’s Docker Manager (e.g., Hostinger’s), go to the Docker Catalog.
- Search for “OpenClaw” and click Deploy.
- In the container setup: Paste your AI API Key, note the auto-generated Gateway Token, and deploy.
The Docker method automatically handles Node.js version compatibility and sandboxing, bypassing manual system requirements.
Access Your Always-Online Dashboard
Once the container status is “running,” open the default port (18789) and access the web dashboard using your Gateway Token. Your AI agent runtime is now live.
A Dockerized VPS install is the gold standard. It provides isolation, easier backups, and eliminates “it works on my machine” problems.
Essential Post-Install Commands for Control & Diagnostics
Once installed, manage your gateway with these key terminal commands.
openclaw: Launches the Text User Interface (TUI) for visual management.openclaw status: Checks the current gateway operational status.openclaw health: Runs a full health diagnostic on all services.openclaw doctor&openclaw doctor fix: Scans for issues and can auto-repair common problems.openclaw update --channel stable: Safely updates your installation to the latest stable release.
Security Best Practices: Sandboxing is Your First Priority
OpenClaw agents can execute tasks with significant access. Terminal access is dangerous, so sandboxing is not optional.
- Use Docker Sandboxing: The safest method. In config, set
agents.defaults.sandbox.mode = "non-main"so only the main agent has host access; others are isolated. - Employ Strong AI Models: Use capable models like Claude 3.5 Sonnet or GPT-4o for reliable and safe decision-making.
- Run Deep Audits: Regularly execute
openclaw security-audit --deepto uncover vulnerabilities. - Secure Integrations: For Telegram, never share the BotFather HTTP API token. For WhatsApp, always whitelist your number in
channels.whatsapp.allowFrom.
Never run OpenClaw in an unsandboxed environment on your primary machine. Always assume an agent could be tricked into running a harmful command.
