How to Install OpenClaw on a VPS in 10 Minutes (2026)

TL;DR: Install Node.js 22+, clone OpenClaw, run npx openclaw init, configure your first channel, and you're done. Total time: under 10 minutes for a basic setup with Telegram.

Prerequisites

Before you start, you need:

• A VPS running Ubuntu 22.04 or Debian 12 (minimum 1 vCPU, 2GB RAM)
• SSH access to your VPS
• A Telegram bot token from @BotFather

Step 1: Connect to Your VPS

class="language-bash">ssh root@your-vps-ip

Step 2: Install Node.js

OpenClaw requires Node.js 22 or later. Install it from NodeSource [source]:

class="language-bash">curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
apt install -y nodejs
node -v

Verify the version — you should see v22.x.x or higher.

Step 3: Create a User

Running services as root is bad practice. Create a dedicated user:

class="language-bash">adduser openclaw
usermod -aG sudo openclaw
su - openclaw

Step 4: Install OpenClaw

class="language-bash">npx openclaw init
cd openclaw
npm install

This downloads the latest OpenClaw release and installs dependencies. It takes about 2 minutes on a standard VPS.

Step 5: Configure Your First Channel (Telegram)

Open the config file:

class="language-bash">nano openclaw.yaml

Find the channels section and add your Telegram bot token:

class="language-yaml">channels:
 telegram:
 enabled: true
 botToken: "YOUR_BOT_TOKEN"

Get your bot token from @BotFather on Telegram. Create a new bot, copy the token, paste it in.

Step 6: Configure Your LLM Provider

Add your preferred LLM provider in the providers section:

class="language-yaml">providers:
 openai:
 apiKey: "YOUR_OPENAI_KEY"
 model: gpt-4o-mini

OpenClaw supports OpenAI, Anthropic, Google, Mistral, OpenRouter, Ollama (local), and 15+ other providers.

Step 7: Start OpenClaw

class="language-bash">npx openclaw start

You should see output like:

[2026-05-04 12:00:00] INFO: Starting OpenClaw v2026.4.24
[2026-05-04 12:00:01] INFO: Telegram channel connected
[2026-05-04 12:00:01] INFO: Agent is ready

OpenTelegram, message your bot, and you're talking to your own AI agent.

Running OpenClaw as a Service

To keep your agent running after you close the SSH session, set it up as a systemd service. Create a service file at /etc/systemd/system/openclaw.service with the openclaw user and the working directory set to your OpenClaw installation. Then run systemctl enable openclaw && systemctl start openclaw. The agent will start automatically on boot and restart if it crashes.

For production deployments, consider adding a reverse proxy like Nginx if you plan to expose the web interface, and set up automatic SSL certificates with Certbot. You can also configure log rotation to prevent the agent logs from filling your disk over time — OpenClaw can generate significant log output during active conversations with multiple channels connected simultaneously.

Memory configuration is the next important step after basic setup. Out of the box, OpenClaw uses in-memory storage which resets on restart. For persistent memory, configure LanceDB or SQLite backends — these let your agent remember past conversations, user preferences, and task history across sessions. This turns your agent from a stateless chatbot into a system that accumulates context and improves over time.

What's Next?

• Connect additional channels (Discord, Signal, WhatsApp)
• Configure memory backends (LanceDB, SQLite, wiki)
• Add search plugins (Brave, Tavily, Perplexity)
• Write custom skills for your specific use cases

See the OpenClaw Plugin Directory → for a full list of everything you can enable.

That's it. You now have a self-hosted AI agent running on your own infrastructure. Total time from start to first message: under 10 minutes.

← Back to all posts