Install OpenClaw on Contabo VPS: Step-by-Step Guide

Q: What if I forget my gateway token?

Find it in the OpenClaw config file: `cat ~/openclaw/openclaw.json` and look for the `gateway_token` field. You can also regenerate it by running `openclaw dashboard` from your SSH session.

Installing OpenClaw on a Contabo VPS is one of the most cost-effective ways to run your own self-hosted AI agent — and this guide covers every step of the manual installation process from scratch.

I have done this multiple times and put together the most complete walkthrough I could find. By the end of this guide, you will have OpenClaw installed, connected to Telegram, and accessible via your own custom domain with HTTPS.

What is OpenClaw?

OpenClaw is an open-source AI agent that runs on your own server. You connect it to an AI provider like OpenAI or Anthropic, configure it through an onboarding process, and then interact with it through messaging platforms or a web-based Control UI.

It is not just a chatbot. Depending on what skills and channels you enable, OpenClaw can:

Reply to messages on Telegram, WhatsApp, Discord, or Slack
Monitor repositories and summarize issues for you
Manage files and trigger commands on the server
Run scheduled tasks and automations based on your instructions

The key phrase is runs on your own server. Your data stays within your infrastructure, and you are not dependent on any third-party platform keeping your agent alive. That is where a VPS like Contabo comes in.

Why Contabo for OpenClaw?

Contabo has been around since 2003 and is known for one thing: they offer more raw compute per dollar than almost any other VPS provider in the market. Their plans are aggressively priced while still delivering solid performance — making them one of the best options for running workloads like AI agents where you need consistent uptime and decent RAM without paying a premium.

Here is why I recommend Contabo for OpenClaw:

Generous RAM at low prices — OpenClaw needs at least 2 GB RAM to run comfortably; even Contabo's entry plan gives you 8 GB
Full root SSH access — you get complete control over your server from day one
Stable infrastructure — Contabo runs their own datacenters in Germany, USA, and Asia-Pacific
Ubuntu support — all standard Ubuntu LTS releases are available out of the box
No extra charges — OpenClaw itself is free and open-source

Contabo VPS Plans for OpenClaw

OpenClaw needs at minimum 2 GB of RAM to run comfortably. Contabo's VPS plans all clear this requirement with room to spare.

Spec	VPS S	VPS M	VPS L
vCPU	4 cores	6 cores	8 cores
RAM	8 GB	16 GB	24 GB
NVMe	100 GB	200 GB	300 GB
Traffic	32 TB	32 TB	32 TB
Price	~$6.99/mo	~$12.99/mo	~$19.99/mo

For OpenClaw, the VPS S plan is more than enough to get started. You have 8 GB of RAM, which comfortably handles the agent, the gateway service, and any skills you enable. If you plan to run additional workloads alongside OpenClaw, consider the VPS M.

Get Contabo VPS

What You Need Before Starting

A Contabo VPS running Ubuntu (20.04 or 22.04 LTS recommended)
An OpenAI API key (or another supported AI provider key)
A Telegram bot token from @BotFather on Telegram
SSH access to your server (terminal on Mac/Linux, or Windows Terminal / PuTTY on Windows)
A domain or subdomain pointed at your VPS (needed for the dashboard — covered in Step 15)

How to Install OpenClaw on Contabo VPS: Step by Step

Here is a quick checklist of every step covered in this guide. Use it as a reference to track your progress:

Step 1 — Buy a VPS with Ubuntu

Go to Contabo VPS and select your plan. During the order process:

Under Operating System, choose Ubuntu — either 20.04 LTS or 22.04 LTS
Set a strong root password — you will need this to SSH into your server
Complete your order

Once your order is confirmed, you will receive your VPS IP address and credentials by email. You can also find them in the Contabo Customer Control Panel.

Step 2 — Log In to Your VPS via SSH

Open a terminal on your computer and connect to your VPS:

ssh root@<your-server-ip>

Replace <your-server-ip> with the IP address from your Contabo dashboard. Enter your root password when prompted.

You are now inside your VPS via a live terminal session.

New to SSH? On Windows, you can use Windows Terminal or PuTTY. On Mac and Linux, the terminal app works out of the box.

Step 3 — Install OpenClaw (Without Onboarding)

Run the following command to install OpenClaw without starting the onboarding wizard immediately. This lets you gather everything you need first before configuring.

curl -fsSL https://openclaw.ai/install.sh | bash -s -- --no-onboard

This command downloads and installs the OpenClaw CLI and gateway service on your server. The --skip-onboard flag prevents the onboarding wizard from launching automatically so you can complete the prerequisites first.

Wait for the installation to finish. You will see a confirmation message when it completes.

Step 4 — Get Your OpenAI API Key

Before running onboarding, you need an API key from your AI provider. This guide uses OpenAI as the example.

Go to platform.openai.com
Create a new Project called "OpenClaw" (keeps usage isolated and easy to revoke)
Go to API Keys inside that project and click Create new secret key
Copy the key and save it somewhere safe — you will need it in the next step

You are responsible for your API usage costs. For light personal use, costs are typically under $5/month. Set a spending limit in your OpenAI dashboard to stay in control.

Step 5 — Get a Telegram Bot Token

OpenClaw connects to Telegram via a bot. You need to create one through BotFather:

Open Telegram and search for @BotFather
Send the command /newbot
Choose a name for your bot (e.g., "My OpenClaw Bot")
Choose a username ending in bot (e.g., myopenclaw_bot)
BotFather will reply with your bot token — it looks like 123456789:ABCDefgh...

Copy this token and keep it ready for the onboarding step.

Step 6 — Run the Onboarding Wizard

Back in your SSH terminal, start the OpenClaw onboarding process:

openclaw onboard

This launches the interactive setup wizard that walks you through configuring OpenClaw for the first time.

Step 7 — Complete the Onboarding

The onboarding wizard will guide you through several steps. Here is what to expect:

Security acknowledgment

OpenClaw will display a security warning. Read it and confirm — this is standard. It is informing you that OpenClaw can read files and run actions on your server when certain skills are enabled.

Choose onboarding mode

Select QuickStart unless you want to configure everything manually from the start.

Select your AI provider and enter your API key

Choose OpenAI (or your preferred provider) and paste the API key you generated in Step 4.

Connect Telegram

When prompted for a messaging channel, select Telegram and paste the bot token you got from BotFather in Step 5.

Skill configuration

You can skip all skills for now. Enable them later from the Control UI once everything is working.

Once onboarding is done, OpenClaw will install and start the gateway service, initialize the agent, and confirm that everything is running.

Step 8 — Message the Bot /start on Telegram

Open Telegram and find the bot you just created (search by its username). Send it the command:

/start

The bot will respond with a pairing command — a special command that links your Telegram account to your OpenClaw instance. Copy this command exactly as shown.

Step 9 — Paste the Pairing Command in Your SSH Terminal

Back in your SSH session, paste the pairing command you received from the Telegram bot and run it. It will look something like:

openclaw pairing approve telegram <pairing-token>

Run the command. OpenClaw will confirm that your Telegram account is now paired and authorized to interact with the agent.

Step 10 — You Are Ready to Use OpenClaw

At this point, OpenClaw is fully installed and operational on your Contabo VPS. You can:

Send messages to your bot on Telegram and get AI responses
Ask it to run tasks, answer questions, or perform actions you have enabled

The next steps cover how to find your gateway token, test the installation, and access the web dashboard.

Step 11 — Find Your Gateway Token

The gateway token is the key you use to log into the OpenClaw Control UI from a browser. It is stored on your server in the OpenClaw configuration file.

SSH into your server and run:

cat ~/openclaw/openclaw.json

Look for the gateway_token field in the JSON output. Copy its value — you will need this to log in to the dashboard.

Step 12 — Test with the OpenClaw Doctor Command

Run the built-in health check to verify that all components are working correctly:

openclaw doctor

This command checks:

That the OpenClaw service is running
That your AI provider connection is valid
That your configured channels (Telegram, etc.) are reachable
That the gateway service is up

If anything is not working, the doctor output will tell you what to investigate.

Step 13 — Check Gateway Status

To confirm the gateway service is running and ready to serve the Control UI:

openclaw gateway status

You should see a status message confirming the gateway is active and listening. If it is not running, start it with:

openclaw gateway start

Step 14 — Create an SSH Tunnel for Access

The Control UI runs on localhost port 18789 on your server. To access it in your browser without a domain, create an SSH tunnel.

Open a new terminal window (keep your existing SSH session alive) and run:

ssh -N -L 18789:127.0.0.1:18789 root@<your-server-ip>

Keep this terminal open. Now open your browser and go to:

http://localhost:18789/

You will see the OpenClaw login page. Paste your gateway token from Step 11 and click Login.

This is the quick-access method. For permanent access without running this command every time, continue to the next steps to set up a custom domain.

Step 15 — Configure Your Domain DNS

To access the OpenClaw dashboard from anywhere via a custom domain, first point a subdomain at your Contabo VPS.

In your DNS manager (Cloudflare, Namecheap, etc.):

Create a new A record
Name: something like ai or openclaw (this creates ai.yourdomain.com)
Value: your Contabo VPS IP address
TTL: Auto or 300 seconds

Wait a few minutes for DNS to propagate before proceeding. You can check propagation at dnschecker.org.

Step 16 — Install Caddy and Configure the Reverse Proxy

Caddy is a lightweight web server that handles HTTPS certificates automatically. Install it and configure it to proxy requests to OpenClaw.

Install Caddy:

apt update && apt install -y caddy

Edit the Caddy configuration:

nano /etc/caddy/Caddyfile

Replace the contents with the following (substitute your actual subdomain):

ai.yourdomain.com {
    reverse_proxy localhost:18789
}

Save with Ctrl+X → Y → Enter, then restart Caddy:

systemctl restart caddy

Caddy will automatically obtain and renew an SSL certificate for your subdomain. After a moment, your Control UI will be accessible at https://ai.yourdomain.com.

Step 17 — Add Your Domain to OpenClaw's allowedOrigins

If you try to access the dashboard and see an error about the origin not being allowed, you need to whitelist your domain in the OpenClaw configuration.

Open the OpenClaw config file:

nano ~/.openclaw/openclaw.json

Find the control_ui section and add your domain to the allowedOrigins array:

"controlUi": {
  "allowedOrigins": [
    "https://ai.yourdomain.com"
  ]
}

Save the file, then restart the OpenClaw gateway for the change to take effect:

openclaw gateway restart

Now try accessing your dashboard again — the origin error should be resolved.

Step 18 — Approve Pending Device Pairing

If the pairing did not complete automatically, OpenClaw will move the device to a pending state. You will need to approve it manually from your SSH terminal.

First, list all pending devices to find the device ID:

openclaw devices list

You will see an output listing paired and pending devices, each with a UUID like 27610061-6eaa-452e-ae77-072d4542c2b7. Copy the UUID of the device showing as pending, then approve it:

openclaw devices approve <device-id>

For example:

openclaw devices approve 27610061-6eaa-452e-ae77-072d4542c2b7

Once approved, the device is authorized and pairing is complete.

Step 19 — Access the OpenClaw Dashboard via Your Custom Domain

Open your browser and go to:

https://ai.yourdomain.com

Log in with your gateway token from Step 11. You now have permanent, HTTPS-secured access to the full OpenClaw Control UI — from anywhere, without needing an SSH tunnel.

From the dashboard you can:

Manage channels (Telegram, WhatsApp, Discord, Slack)
Enable and configure skills
Review agent activity and logs
Update your AI provider settings
Manage devices and pairing

Managing Your AI API Key and Costs

OpenClaw interacts with third-party AI providers like OpenAI, which means API usage costs apply based on how much your agent processes. Here is how to keep things under control:

Set usage limits in your OpenAI dashboard under Billing → Usage limits
Create a dedicated project for OpenClaw so you can isolate its usage and revoke the key independently
Monitor your spending monthly — for light personal use, costs are typically under $5/month

You are fully responsible for your API usage costs. Contabo has no visibility into or control over your OpenAI or Anthropic usage.

Contabo vs Hostinger for OpenClaw: Which Should You Choose?

Both are excellent options for self-hosting OpenClaw. Here is how they compare:

	Contabo	Hostinger
Install method	Manual via SSH (this guide)	Docker Manager (browser-based)
Onboarding	SSH wizard after install	Web UI from day one
Control UI access	SSH tunnel or custom domain with Caddy	Accessible via Docker port mapping
Free AI credits	None included	10 free credits (~$10 value)
Pricing	Lower cost per spec	Slightly higher, more beginner-friendly
Best for	Power users comfortable with SSH	Beginners, non-technical users

My recommendation: If you are comfortable with SSH and want the most raw compute for your money, go with Contabo. If you prefer a GUI-first experience and want Hostinger's 10 free AI credits to test before committing to a paid API plan, Hostinger VPS is the easier starting point.

Already on Hostinger? I also have a detailed guide on how to install OpenClaw on Hostinger VPS which covers the Docker Manager one-click deploy approach.

Important Notes and Security Tips

OpenClaw can read files and run actions on your server. Only enable skills you understand and trust.
Keep your gateway token private — treat it like a password. Anyone with the token can access your Control UI.
Use dedicated API keys for OpenClaw — not your master account key. Rotate keys periodically.
Enable SSH key authentication instead of password-only login for better server security.
Review OpenClaw's skill permissions before enabling anything that accesses external services.

FAQs

What Ubuntu version should I use for OpenClaw?

Ubuntu 22.04 LTS is recommended. Ubuntu 20.04 LTS also works. Avoid non-LTS releases for a production self-hosted agent.

What AI providers work with OpenClaw on Contabo?

OpenClaw supports any AI provider that exposes an API — including OpenAI (GPT-4, GPT-4o), Anthropic (Claude), Google Gemini, Mistral, and others. You select your provider and enter your API key during the onboarding wizard.

Can I use OpenClaw without a domain name?

Yes. You can always access the Control UI via SSH tunnel using ssh -N -L 18789:127.0.0.1:18789 root@<ip>. A domain is optional but recommended for permanent access without running the tunnel command each time.

What if I forget my gateway token?

Find it in the OpenClaw config file: cat ~/openclaw/openclaw.json and look for the gateway_token field. You can also regenerate it by running openclaw dashboard from your SSH session.

Can I connect multiple messaging channels at once?

Yes. From the Control UI, you can add Telegram, WhatsApp, Discord, Slack, and other supported channels simultaneously. Each channel can trigger the same agent.

How do I update OpenClaw after installation?

Refer to the official OpenClaw documentation for the update process — it is managed through the OpenClaw CLI on your server.

Can I run OpenClaw alongside other applications on the same VPS?

Yes. With 8 GB of RAM on Contabo's VPS S, there is plenty of headroom to run OpenClaw alongside other lightweight services. Just be mindful of port conflicts and resource usage if you stack multiple applications.

Final Thoughts

The manual installation approach gives you full control over every part of your OpenClaw setup — from how the agent is configured to how you access the dashboard. It takes a bit more work than a one-click deploy, but the result is a cleaner, more transparent installation that you understand end to end.

If you are ready to run your own AI agent on infrastructure you actually own and control, Contabo VPS is one of the best value options available right now — generous specs at prices that are hard to beat anywhere else.