How to Use OpenClaw in Windows: Complete Setup Guide

OpenClaw (formerly known as Clawdbot and briefly Moltbot) is your personal, open-source AI assistant that can automate tasks, manage your workflow, and act as your digital personal assistant. If you're on Windows, you'll need a specific setup approach to get it working smoothly. This guide walks you through the complete installation and configuration process.

What is OpenClaw?

OpenClaw is an autonomous AI personal assistant software built by Peter Steinberger and the community. It operates as an intelligent agent capable of executing tasks across multiple platforms including WhatsApp, Telegram, Discord, and more. You can deploy it locally or on private servers, and it integrates with popular AI models like Anthropic's Claude and OpenAI's models.

The project has grown tremendously since its 2025 release, becoming one of the fastest-growing GitHub repositories. It's now the go-to choice for users who want a customizable, privacy-focused AI assistant.

Prerequisites Before You Start

Before installing OpenClaw on Windows, make sure you have:

Essential Requirements:

Node.js version 22 or higher (download from nodejs.org)
Windows 10 or Windows 11
Administrator access to your machine
A stable internet connection

Highly Recommended:

API key from Anthropic (for Claude models) — this is the recommended authentication method
Alternatively, an OpenAI API key or OAuth credentials
Brave Search API key for web search functionality (optional but useful)

Important Note for Windows Users: OpenClaw on Windows requires WSL2 (Windows Subsystem for Linux 2). Native Windows installation is untested and has compatibility issues. WSL2 is strongly recommended with Ubuntu as the Linux distribution.

Step-by-Step Installation Guide

Step 1: Install WSL2 and Ubuntu

Since OpenClaw requires WSL2, your first step is setting up the Windows Subsystem for Linux.

On Windows 11:

Open PowerShell as Administrator and run:

wsl --install

This automatically installs WSL2 and Ubuntu. Once complete, restart your computer.

On Windows 10:

Follow the official WSL2 installation guide from Microsoft.

After installation, launch Ubuntu and set up a username and password for your Linux account.

Step 2: Verify Node.js Installation

Inside your Ubuntu terminal (WSL2), check if Node.js is available:

node --version
npm --version

If Node.js isn't installed, install it using apt:

sudo apt update
sudo apt install nodejs npm

Verify you have Node 22 or higher. If you need a newer version, use Node Version Manager (nvm):

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.0/install.sh | bash
nvm install 22
nvm use 22

Step 3: Install OpenClaw CLI

In your WSL2 Ubuntu terminal, run the OpenClaw installation script:

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

This downloads and installs the OpenClaw command-line interface globally. Once the script completes, verify the installation:

openclaw --version

Step 4: Run the Onboarding Wizard

The onboarding wizard is the recommended setup path. It walks you through configuration interactively:

openclaw onboard --install-daemon

During the wizard, you'll make these choices:

A) Local vs Remote Gateway

Choose "Local" for personal use on your machine
Choose "Remote" if you're setting up on a server

B) Authentication Method

Recommended: Use an Anthropic API key (get it from console.anthropic.com)
Alternative: Use OpenAI API key for GPT models
Legacy: OAuth credentials (requires a browser login)

C) Chat Channels

Select where you want to access your AI assistant:

WhatsApp (QR code login recommended)
Telegram (requires bot token)
Discord (requires bot token)
Slack, Google Chat, Microsoft Teams, Signal, etc.

You can set up multiple channels, but start with one if you're new to OpenClaw.

D) Daemon Installation

Enable "Install daemon" to automatically start OpenClaw as a background service using systemd in WSL2.

E) Gateway Configuration

The wizard generates a gateway token automatically. This token authenticates connections to your OpenClaw instance.

Step 5: Configure Your API Key

If you chose Anthropic during onboarding, store your API key:

openclaw configure --section web

This command saves your credentials securely in ~/.openclaw/agents/[agentId]/agent/auth-profiles.json.

Important Security Note: For headless/server setups, run OAuth authentication on your regular Windows machine first, then copy the credentials to your WSL2 instance.

Step 6: Start the Gateway

If you installed the daemon during onboarding, the Gateway should start automatically. Check its status:

openclaw gateway status

To manually start the Gateway in the foreground:

openclaw gateway --port 18789 --verbose

The Gateway is OpenClaw's core service that connects all your chat channels and manages AI interactions.

Step 7: Access the Dashboard

OpenClaw provides a web-based dashboard for quick testing and management. Open your browser and navigate to:

http://127.0.0.1:18789/

If prompted, enter the gateway token generated during onboarding. This dashboard lets you chat directly with your AI assistant without setting up external channels.

Step 8: Connect Your First Chat Channel

For WhatsApp (Recommended):

openclaw channels login

Scan the QR code using WhatsApp → Settings → Linked Devices. Your account will then be connected.

For Telegram:

Create a bot using @BotFather on Telegram
Get your bot token
Configure it: openclaw configure --section channels.telegram
Message your bot — it will return a pairing code

For Discord:

Create an application on Discord Developer Portal
Get your bot token
Configure and start receiving messages

Step 9: Verify Everything Works

Run the health check:

openclaw health
openclaw status

Send a test message:

openclaw message send --target +1234567890 --message "Hello from OpenClaw"

(Replace with your actual phone number for WhatsApp, or your Telegram/Discord ID)

Important Windows-Specific Considerations

Runtime Warning

Bun is not recommended for Windows/WSL2 users, especially if you use WhatsApp or Telegram channels. Bun has known compatibility issues with these channels. Always use Node.js.

Performance Tips

WSL2 runs Linux virtualization, so performance is good but slightly slower than native Windows
For always-on setups, consider running OpenClaw on a VPS or cloud server instead
Ensure adequate RAM (4GB minimum, 8GB recommended)

File System Access

Your OpenClaw configuration lives in your WSL2 Ubuntu home directory:

~/.openclaw/

From Windows File Explorer, access this location via:

\\wsl$\Ubuntu\home\[username]\.openclaw\

Common Setup Scenarios

Scenario 1: Personal Assistant via WhatsApp

This is the simplest setup for personal use. After onboarding with WhatsApp channel selected, you can start sending messages to your OpenClaw assistant from your phone.

Scenario 2: Team Bot on Telegram

Set up Telegram during onboarding, and your team can interact with the bot in group chats. Configure channel routing to send specific types of messages to different agents if needed.

Scenario 3: Always-On Server

For production use, deploy OpenClaw on a VPS (like Render, Railway, or Northflank) instead of your personal Windows machine. Use Tailscale or SSH tunnels for secure remote access.

Security Best Practices

Store credentials safely: Never hardcode API keys. Use the configuration system to store them securely.
Use pairing approvals: By default, unknown DMs require approval before the assistant responds. This prevents spam.
Limit exposed features: Review which tools and integrations you enable. Each one increases the attack surface.
Run in sandbox mode: Use agents.defaults.sandbox.mode: "non-main" to isolate agent operations.
Regular security audits: Run openclaw security audit --deep to identify potential vulnerabilities.

Next Steps After Installation

Once OpenClaw is running smoothly:

Add more channels: Integrate with Slack, Discord, Telegram groups, etc.
Configure skills: Enable specific tools like web search, email integration, or calendar management
Set up automation: Use cron jobs or webhooks to trigger tasks automatically
Customize prompts: Modify system prompts to match your preferences
Add workspace skills: Install additional skills from the community (ClawHub)

Troubleshooting Common Issues

Gateway won't start:

Check if Node 22+ is installed: node --version
Try restarting WSL2: Exit Ubuntu terminal and run wsl --shutdown in PowerShell
Check gateway logs: openclaw logs gateway

Channel connection fails:

Verify API credentials: openclaw status --all
Re-run onboarding for the specific channel: openclaw configure --section channels.telegram
Check internet connectivity: ping -c 1 google.com

Permission errors in WSL2:

Ensure you're running terminal as your user, not root
Check file permissions: ls -la ~/.openclaw/

Slow response times:

Check WSL2 system resources: free -h
Reduce simultaneous agent tasks or increase allocated RAM

FAQ: Your Questions Answered

Q: Do I need to run OpenClaw 24/7?

A: It depends on your use case. For WhatsApp/Telegram channels, the daemon will keep the gateway running in the background as a systemd service. If you only want to use the web dashboard occasionally, you can manually start and stop the gateway. For production use or always-on assistants, deploy on a cloud server instead of your Windows machine.

Q: Can I run OpenClaw on native Windows without WSL2?

A: Native Windows installation is untested and has significant compatibility issues. WSL2 is strongly recommended and not difficult to set up. Using WSL2 ensures better tool compatibility and prevents many headaches.

Q: Which AI model should I use?

A: For long-context strength and better prompt-injection resistance, use Anthropic Pro/Max with Claude Opus 4.5. However, you can use any model supported by OpenClaw. If cost is a concern, Claude Haiku 4.5 is faster and cheaper but less capable for complex tasks.

Q: Is OpenClaw safe to use?

A: OpenClaw is open-source and can be reviewed by security researchers. However, it requires broad permissions to function (file system, internet access, chat channels, etc.). Security researchers have noted potential risks with prompt injection and exposed administrative interfaces. It's best suited for advanced users who understand the security implications. Always run security audits before connecting to production systems or accounts with sensitive credentials.

Q: Can I use multiple AI providers (OpenAI + Anthropic)?

A: Yes. OpenClaw supports multiple model providers and can failover between them if one is unavailable. Configure different providers in your auth profiles and specify which to use for different tasks.

Q: How do I update OpenClaw?

A: Inside WSL2 Ubuntu, run:

openclaw update

# Or reinstall from npm:
npm install -g openclaw@latest

Q: Can I access my Windows files from OpenClaw?

A: Yes. WSL2 can access your Windows file system via /mnt/c/Users/[username]/. You can configure OpenClaw to interact with files in specific directories. Be careful about granting broad file system access for security reasons.

Q: What if I already have Node.js installed on Windows (not WSL2)?

A: OpenClaw won't work properly with Node.js installed only on Windows. You need Node.js inside WSL2/Ubuntu. Install it there as described in Step 2, or ensure the Windows Node.js is accessible from WSL2 (advanced).

Q: Can I run multiple OpenClaw instances?

A: Yes, you can run multiple instances with different configurations on different ports. Configure separate agents, workspaces, and gateway instances as needed.

Q: How much disk space does OpenClaw need?

A: Base installation is ~500MB. Configuration, logs, and cached data can add another 1-2GB depending on usage. The bulk of space is taken by node_modules and browser automation tools if enabled.

Q: Can I back up my OpenClaw configuration?

A: Yes. Back up the entire ~/.openclaw/ directory. This includes all your credentials, configurations, workspace settings, and logs. Store backups securely.

Q: Does OpenClaw work offline?

A: The gateway can run locally without internet, but you still need internet for API calls to AI models (Claude, GPT, etc.) and for accessing external services through integrations. Some tools like file system access and command execution can work offline.

Q: What are "skills" in OpenClaw?

A: Skills are modular plugins that extend OpenClaw's capabilities. Built-in skills include web search, email, calendar management, and command execution. You can also build custom skills or install community skills from ClawHub.

Q: Is there a macOS or Linux version?

A: Yes. OpenClaw has native support for macOS (with a dedicated menu bar app) and Linux. The Windows version specifically requires WSL2 because native Windows compatibility is limited.

Q: How do I uninstall OpenClaw?

A: Run:

openclaw uninstall

# Or manually remove:
npm uninstall -g openclaw
rm -rf ~/.openclaw/

Q: Can I contribute to OpenClaw?

A: Absolutely! OpenClaw is open-source on GitHub. The community actively develops new features and skills. Check the CONTRIBUTING.md file on GitHub for guidelines.

Final Thoughts

OpenClaw is a powerful tool for creating your own autonomous AI assistant. While the Windows setup requires WSL2, the process is straightforward and worth the effort. The flexibility to connect across WhatsApp, Telegram, Discord, Slack, and more makes it unique.

Start with the basic setup, get comfortable with the dashboard, and then gradually explore advanced features like custom skills, automation hooks, and multi-agent routing.

Happy automating! 🦞

Version: Based on OpenClaw documentation as of February 2026
Need Help? Check the official OpenClaw docs or visit the GitHub repository at github.com/openclaw/openclaw