How to Install OpenClaw: Complete Setup Guide for Mac, Windows, and Linux (2026)
Paul TherbieoHow to Install OpenClaw
OpenClaw runs on any operating system that supports Node.js 22 or higher. The full installation, from downloading Node.js to sending your first message, takes about 15 to 20 minutes. This guide walks through every step.
For a Docker-based install instead, see the OpenClaw Docker guide. For always-on operation on a Mac Mini, see the Mac Mini setup guide.
Requirements
Before you start, make sure you have:
- Node.js 22 or higher (check your current version with
node --version) - An API key from a supported model provider: Anthropic, OpenAI, Google, or a local model via Ollama
- Administrator access on your machine for the daemon installation step
If Node.js is not installed, download it from nodejs.org and choose the latest LTS release.
Step 1: Install OpenClaw via npm
Open a terminal and run:
npm install -g openclaw@latest
This installs the openclaw CLI globally on your system. Confirm it worked:
openclaw --version
Step 2: Run the Onboarding Wizard
The onboarding wizard sets up your configuration directory at ~/.openclaw/ and installs the gateway as a persistent background service:
openclaw onboard --install-daemon
The --install-daemon flag installs OpenClaw as a system service. On macOS, this creates a launchd daemon. On Linux, a systemd service. After this step, the gateway starts automatically every time your machine boots without any manual action.
The wizard will prompt you for your preferred LLM provider and API key, plus a few basic preferences for the agent's behavior.
Step 3: Configure Your Model
Edit ~/.openclaw/openclaw.json to set the model you want to use:
{
"agent": {
"model": "anthropic/claude-opus-4-6"
}
}
Supported model formats:
| Provider | Config string |
|---|---|
| Anthropic Claude | anthropic/claude-opus-4-6 |
| OpenAI GPT-4 | openai/gpt-4 |
| Google Gemini | google/gemini-3-pro |
| Local via Ollama | ollama/llama3.3 |
If you use Ollama for local models, start Ollama separately with ollama serve before starting the OpenClaw gateway. Local models require no API key and have no per-token cost.
Step 4: Connect Your Messaging Channels
To connect WhatsApp, Telegram, Discord, Slack, Signal, iMessage, Microsoft Teams, or another supported channel, run:
openclaw channels login
This opens a browser window where you authenticate with each service. You only need to do this once per channel. After that, the connection persists across restarts.
If you want to test OpenClaw without connecting a messaging app first, skip this step and use the built-in WebChat in the Control UI.
Step 5: Start the Gateway
Launch the OpenClaw gateway for the first time:
openclaw gateway --port 18789
Open http://127.0.0.1:18789/ in your browser. This is the Control UI where you monitor the agent, review memory, manage installed skills, and adjust configuration.
After the initial start, the daemon handles all future startups automatically.
Step 6: Send Your First Message
If you connected a channel in Step 4, open that app and send the agent a message:
Hello! What can you do?
The agent responds with a summary of its current capabilities based on installed skills.
If you are using WebChat, click the chat icon in the Control UI at http://127.0.0.1:18789/.
Updating OpenClaw
To update to the latest stable version:
openclaw update --channel stable
OpenClaw also supports beta and dev channels for users who want early access to new features before the stable release.
Common Installation Problems
"node: command not found" Node.js is not installed or is not in your PATH. Download version 22 or higher from nodejs.org.
"EACCES: permission denied" during npm install
Either run with sudo or configure npm to install global packages in a directory you own. See the npm docs on fixing permissions.
Gateway fails to start
Check whether port 18789 is already in use: lsof -i :18789. Use --port to choose a different port if needed.
Channel authentication loops
Try logging out and back in: openclaw channels logout && openclaw channels login.
Daemon does not start after reboot
Run openclaw onboard --install-daemon again to re-register the daemon. If the problem persists, check your system's service logs (Console.app on macOS, journalctl on Linux).
Next Steps
Once OpenClaw is running, install skills to extend what it can do. The ClawHub registry has over 5,700 skills. Read the complete skills guide to learn how to find and install them.
For crypto and trading automation, see the trading skills guide.
Frequently Asked Questions
How long does OpenClaw installation take?
About 15 to 20 minutes for a complete setup including channel connections. The npm install itself takes under a minute.
Can I install OpenClaw without administrator access?
Yes. The daemon installation requires elevated permissions, but you can run OpenClaw manually with openclaw gateway without installing the daemon. You will need to start it manually after each reboot.
Does OpenClaw work on Windows?
Yes. OpenClaw supports Mac, Windows, and Linux. On Windows, the daemon uses a comparable service mechanism to launchd (macOS) and systemd (Linux).
What is the minimum Node.js version?
Node.js 22. Earlier versions are not supported and will produce errors.
Can I run multiple OpenClaw instances on the same machine?
Yes, on different ports. Use --port to specify a unique port for each instance.