How to Install OpenClaw: Step-by-Step Tutorial

Getting OpenClaw running is surprisingly simple. The entire platform is a Node.js application — no GPU required, no complex dependencies, no Docker wizardry (unless you want it). In this guide, I'll walk through every installation method so you can pick the one that matches your setup.

Prerequisites

Before you start, make sure you have:

• Node.js 18 or higher — Check with node --version
• A machine to run the gateway — Mac, Linux, or Windows (WSL). A Mac Mini, Linux VPS, or even a Raspberry Pi works.
• An AI model API key — You'll need at least one: OpenAI, Anthropic, or another supported provider

That's it. No GPU, no CUDA, no 32GB of RAM. The gateway is lightweight — it orchestrates, it doesn't compute. The heavy lifting happens on the AI provider's servers.

Method 1: Homebrew (macOS — Recommended)

If you're on a Mac, this is the fastest path:

That's genuinely it. Homebrew handles dependencies, creates the workspace directory at ~/.openclaw/workspace/, and sets up the CLI. The gateway starts as a background daemon.

Method 2: npm (Cross-Platform)

Works on macOS, Linux, and Windows (via WSL):

The npm method is the most portable. If you have Node.js, you have OpenClaw. This is also the recommended approach for VPS deployments.

Method 3: Docker

If you prefer containerized deployments:

Docker is ideal for users who want isolation or are deploying alongside other services. The mounted volume ensures your workspace files persist between container restarts.

Method 4: Manual / From Source

For developers who want to inspect or modify the code:

Building from source gives you full access to the codebase. This is the MIT-licensed, open-source project — you can fork it, modify it, contribute to it, or just read the code to understand what's happening under the hood.

Post-Installation Setup

Once the gateway is running, you need to configure three things:

1. Set Up Your AI Model

OpenClaw needs at least one AI model to function. The most common choices:

You can use any combination of providers. Many users set up multiple models and route different tasks to different models based on cost and capability.

2. Connect Your First Channel

Your agent needs somewhere to talk. Telegram is the easiest first channel:

For other channels, see our WhatsApp guide or the complete guide.

3. Create Your Workspace Files

The workspace at ~/.openclaw/workspace/ defines your agent. At minimum, create these files:

These files are just the starting point. Read our SOUL.md deep dive and Building Your First Agent tutorial for complete guidance on crafting these files.

Verify Everything Works

Let's confirm your installation:

Now open Telegram, find your bot, and send it a message. If everything is configured correctly, your agent will respond. Congratulations — you have a personal AI agent running on your hardware.

Keeping OpenClaw Running 24/7

For the agent to be always available, the gateway needs to stay running. Here are the common approaches:

macOS: launchd (Automatic)

The Homebrew installation typically sets up a launchd service automatically. If not:

Linux/VPS: pm2 (Recommended)

pm2 is a Node.js process manager that handles restarts, logging, and monitoring:

Linux/VPS: systemd

For systemd-based Linux distributions:

Common Installation Issues

Node.js Version Too Old

OpenClaw requires Node.js 18+. If you're on an older version:

Permission Errors on npm Global Install

If you get EACCES errors:

Gateway Won't Start

Check the logs:

Channel Connection Fails

Most channel issues are authentication-related. Double-check your bot token (Telegram) or phone number linking (WhatsApp). Our Telegram and WhatsApp guides cover troubleshooting in detail.

Recommended VPS Providers

If you don't have a Mac Mini or Linux server at home, a VPS works perfectly:

• Hetzner — €3.99/mo for 2 vCPU, 4GB RAM. Best value in Europe.
• DigitalOcean — $6/mo for 1 vCPU, 1GB RAM. Simple, reliable.
• Linode (Akamai) — $5/mo for 1 vCPU, 1GB RAM. Good documentation.
• AWS Lightsail — $3.50/mo for 512MB RAM. Cheapest option.

Any of these will run OpenClaw comfortably. The gateway is lightweight — most of the CPU and memory usage goes to the Node.js runtime itself, not the OpenClaw application.

Next Steps

You've got OpenClaw installed and running. Here's your roadmap:

1. Craft Your SOUL.md — Give your agent a real personality
2. Build Your First Agent — Complete workspace configuration
3. Set Up Memory — Enable persistent context
4. Add Skills — Web search, file access, browser control
5. Enable Automation — Morning briefings, scheduled tasks

Frequently Asked Questions