02 - Installer OpenClaw sur un VPS en 30 minutes

Introduction

You've discovered what OpenClaw is and are convinced? Perfect. In this step-by-step guide, we'll install OpenClaw on a VPS (Virtual Private Server) from scratch. In 30 minutes, you'll have a functional AI agent connected to Telegram, ready to work for you 24/7.

No unnecessary chatter — just concrete steps.

Step 1: Choose Your VPS

Recommended Options

💡 Recommendation for beginners : Hostinger VPS KVM 1 (1 vCPU, 4 GB RAM, 50 GB SSD) for ~€4/month. With our partner code, you'll get 20% off on an annual package. Clear interface, French support, perfect for beginners. Alternative: Hetzner CX22 for advanced users.

Minimum Required Configuration

• CPU: 1 vCPU (sufficient)
• RAM: 2 GB minimum (4 GB recommended)
• Disk: 20 GB minimum
• OS: Ubuntu 22.04 or 24.04 LTS (recommended)
• Network: Public IPv4

⚠️ Important: OpenClaw itself consumes few resources. The Gateway is a lightweight Node.js process. AI computing power is outsourced to APIs (OpenRouter, Anthropic, etc.).

Step 2: Connect via SSH

Once your VPS is created, connect to it:

ssh [[username]]@[[ip]]

If you've configured an SSH key (recommended):

ssh -i ~/.ssh/my_key [[username]]@[[ip]]

First thing to do — update the system:

apt update && apt upgrade -y

Basic Security (optional but recommended)

# Create a non-root user
adduser openclaw
usermod -aG sudo openclaw

# Configure the firewall
ufw allow 22/tcp    # SSH
ufw allow 443/tcp   # HTTPS (if you expose the web interface)
ufw enable

💡 Tip: For a complete security guide, check out our article Securing OpenClaw.

Step 3: Install Node.js 22+

OpenClaw requires Node.js version 22 or higher. Here's how to install it properly:

Recommended Method (NodeSource)

# Install Node.js 22 via NodeSource
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo bash -
apt install -y nodejs

Verify the installation:

node --version   # Should display v22.x.x or higher
npm --version    # Should display 10.x.x or higher

Alternative Method (nvm)

If you prefer managing multiple Node versions:

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.0/install.sh | bash
source ~/.bashrc
nvm install 22
nvm use 22

⚠️ Warning: If you use nvm, ensure that the systemd service (which we'll configure later) points to the correct Node binary.

Step 4: Install OpenClaw

Method 1: Installation Script (recommended)

The simplest method — the script handles everything automatically:

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

This script will:
1. Check that Node.js 22+ is installed (and install it if not)
2. Install OpenClaw globally via npm
3. Launch the onboarding assistant

Method 2: Manual Installation via npm

If you prefer controlling each step:

npm install -g openclaw@latest

Then launch the onboarding:

openclaw onboard --install-daemon

💡 In case of error with sharp: If you see compilation errors related to sharp, use:

SHARP_IGNORE_GLOBAL_LIBVIPS=1 npm install -g openclaw@latest

Verify that everything is installed:

openclaw --version

Another useful command for diagnostics:

openclaw doctor

The openclaw doctor command checks your installation and reports any issues. It's your best friend for diagnostics.

Step 5: Initial Configuration

Launch Interactive Setup

openclaw setup

This command will:
- Create the configuration file ~/.openclaw/openclaw.json
- Initialize the workspace ~/.openclaw/workspace
- Create base files (SOUL.md, AGENTS.md, USER.md, etc.)

Manual Configuration

If you prefer editing directly, create the configuration file:

mkdir -p ~/.openclaw
nano ~/.openclaw/openclaw.json

Here's a minimal functional configuration:

{
  // Agent workspace
  agents: {
    defaults: {
      workspace: "~/.openclaw/workspace"
    }
  }
}

Step 6: Configure AI Model (OpenRouter)

OpenClaw needs a language model to function. The most flexible solution is OpenRouter, which provides access to dozens of models (Claude, GPT-4, Gemini, Llama…) via a single API.

Create an OpenRouter Account

1. Go to openrouter.ai
2. Create an account
3. Add credits ( $5 is enough to start)
4. Generate an API key in Keys

Configure OpenRouter in OpenClaw

Add the key to your configuration:

openclaw config set models.providers.openrouter.apiKey "sk-or-v1-YOUR_KEY_HERE"

Or edit directly ~/.openclaw/openclaw.json:

{
  agents: {
    defaults: {
      workspace: "~/.openclaw/workspace",
      model: "anthropic/claude-haiku-4-5"  // Default model (fast and economical)
    }
  },
  models: {
    providers: {
      openrouter: {
        apiKey: "sk-or-v1-YOUR_KEY_HERE"
      }
    }
  }
}

💡 Model Choice:
- anthropic/claude-haiku-4-5: Fast, economical (~$0.25/M tokens). Ideal for common tasks.
- anthropic/claude-sonnet-4: More powerful, good quality/price ratio.
- anthropic/claude-opus-4: Most capable, for complex tasks. More expensive.

You can change models on the fly with the /model command in the chat.

Step 7: Connect Telegram

Telegram is the easiest channel to configure with OpenClaw.

Create a Telegram Bot

1. Open Telegram and search for @BotFather
2. Send /newbot
3. Choose a name (e.g., "My AI Agent")
4. Choose a username (e.g., my_ai_agent_bot)
5. BotFather gives you a token — copy it carefully

Configure the Bot in OpenClaw

openclaw config set channels.telegram.token "YOUR_TELEGRAM_TOKEN"

To restrict access to your bot (strongly recommended):

openclaw config set channels.telegram.allowFrom '["YOUR_TELEGRAM_ID"]'

💡 Find Your Telegram ID: Send a message to @userinfobot on Telegram, it will give you your numeric ID.

Or edit the configuration directly:

{
  agents: {
    defaults: {
      workspace: "~/.openclaw/workspace",
      model: "anthropic/claude-haiku-4-5"
    }
  },
  models: {
    providers: {
      openrouter: {
        apiKey: "sk-or-v1-YOUR_KEY_HERE"
      }
    }
  },
  channels: {
    telegram: {
      token: "YOUR_TELEGRAM_TOKEN",
      allowFrom: ["YOUR_TELEGRAM_ID"]
    }
  }
}

⚠️ Security: Without allowFrom, anyone can talk to your bot and potentially execute commands on your server. Always configure this restriction.

Step 8: First Test

Start the Gateway:

openclaw gateway start

Verify that it's running:

openclaw gateway status

To see the overall status:

openclaw health

Now, open Telegram and send a message to your bot. If everything is configured correctly, it should respond!

Quick Test

Send these messages to verify that everything works:

1. "Hello, who are you?" → It should introduce itself
2. "What's the time?" → It should give the server time
3. "Create a file test.txt with 'Hello World'" → It should create the file
4. "Read the file test.txt" → It should display the content

If it works, congratulations 🎉 — your AI agent is operational!

Step 9: systemd Service (Permanent Operation)

To make OpenClaw run permanently and restart automatically:

Automatic Service Installation

openclaw service install

This command creates and enables a systemd service for you.

Verification

To check the service status:

systemctl status openclaw

To see real-time logs via journalctl:

journalctl -u openclaw -f

Alternative via OpenClaw directly:

openclaw logs -f

Useful Commands

To restart the service:

openclaw gateway restart

To stop the service:

openclaw gateway stop

To start the service:

openclaw gateway start

The service will automatically restart in case of a crash and on server reboot.

Step 10: Diagnostics and Troubleshooting

The Magic Command: openclaw doctor

openclaw doctor

This command checks everything:
- Node.js version
- Valid configuration
- Channel connectivity
- Workspace status
- Known issues

If issues are detected:

openclaw doctor --fix

Common Issues

Bot Doesn't Respond on Telegram

1. Check that the Gateway is running: openclaw gateway status
2. Check logs: openclaw logs | tail -50
3. Check Telegram token: is it correct in the config?
4. Check allowFrom: is your ID in it?

"Model not found" Error

1. Check your OpenRouter key: openclaw config get models
2. Check that you have credits on OpenRouter
3. Try another model: /model anthropic/claude-haiku-4-5

Gateway Doesn't Start

To see detailed errors:

openclaw logs --level error

Check configuration:

openclaw doctor

Reset if necessary:

openclaw doctor --fix

openclaw Command Not Found After Installation

If after installation openclaw is not found in a new terminal, check where npm installs global binaries:

npm config get prefix

Add to PATH if necessary:

echo 'export PATH="$(npm config get prefix)/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

Useful Logs

To see complete logs:

openclaw logs

To see real-time logs:

openclaw logs -f

To filter by error level:

openclaw logs --level error

To see warnings:

openclaw logs --level warn

Summary of Commands

Here's a recap of all the commands used:

Installation

Via installation script:

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

Alternative via npm:

npm install -g openclaw@latest

Configuration

To launch interactive setup:

openclaw setup

For onboarding with daemon:

openclaw onboard --install-daemon

To diagnose installation:

openclaw doctor

Gateway Management

Start Gateway:

openclaw gateway start

Stop Gateway:

openclaw gateway stop

Restart Gateway:

openclaw gateway restart

Check status:

openclaw gateway status

systemd Service

Install service:

openclaw service install

Check service status:

systemctl status openclaw

Diagnostics

Diagnose installation:

openclaw doctor

Repair automatically:

openclaw doctor --fix

Check overall health:

openclaw health

See real-time logs:

openclaw logs -f