OpenClaw Tutorial: Getting Started with Your AI Agent on a VPS

So you have heard about OpenClaw. Maybe a friend showed you their AI assistant that books appointments, checks the weather, and responds to DMs while they sleep. Or maybe you stumbled across it on GitHub and thought, "I need that."

Here is the good news: getting started with OpenClaw is easier than you think. This OpenClaw tutorial walks you through everything from picking a host to deploying your first skills. By the end, your AI agent will be alive and working for you.

Let's dig in.

What is OpenClaw?

OpenClaw is an open-source AI agent framework. Think of it as the operating system for your personal AI. It runs on a server (your VPS), connects to messaging platforms like Telegram, Discord, or Slack, and uses large language models like Claude or GPT to understand and act on instructions.

Unlike a chatbot that just talks back, OpenClaw can do things. It reads files, runs commands, searches the web, sets up skills, and even manages its own memory across sessions. It does not just converse with you. It works for you.

The framework is modular by design. You pick the model, you pick the platforms, you pick the skills. The whole thing is configured through a single YAML file. That might sound technical, but you will see in a minute that it is surprisingly approachable.

Why Run OpenClaw on a VPS?

OpenClaw runs wherever Node.js runs. You could run it on your laptop. But if you want your agent to always be on, always listening, and always available, you need a server that stays on. That is where a VPS comes in.

A VPS (Virtual Private Server) gives you a dedicated slice of a server with its own operating system, its own IP address, and full root access. You install whatever you want. It runs 24/7. Your agent never sleeps.

Running OpenClaw on a VPS means:

• Your agent is always online, even when your laptop is closed
• You can connect it to messaging apps that need a persistent webhook
• You can scale up resources if your agent starts doing heavier work
• You have full control over security, updates, and backups

If you want to understand more about why a VPS matters for AI agents, check out our guide on what is an AI agent VPS.

Signing Up and Getting Your VPS

Before you can run OpenClaw, you need a server to run it on. The easiest way is to spin up a VPS that is pre-configured for exactly this kind of workload.

Head over to AgentVPS pricing and pick a plan. For a single OpenClaw instance with Telegram or Discord, the entry-level plan is plenty. You can always upgrade later as your agent gets busier.

After you sign up, you will receive your VPS credentials. This includes an IP address, a username, and either a password or an SSH key. Hold onto those. You will need them in the next step.

Not sure what specs you need? Read our guide on AI agent VPS requirements to understand the tradeoffs between memory, CPU, and storage for different agent workloads.

First Login to Your VPS

You connect to your VPS over SSH. If you are on macOS or Linux, open your terminal. If you are on Windows, you can use PowerShell, WSL, or a tool like PuTTY.

Here is the basic command:

ssh username@your-vps-ip

Type yes if it asks about the host key. Enter your password. You are in.

You should see a command prompt that looks something like:

root@your-vps:~#

You are now logged into your VPS. This is your server. From here, you can install anything, including OpenClaw.

Installing OpenClaw

The OpenClaw project provides a quick install script that handles the heavy lifting. Run this on your VPS:

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

The script checks for Node.js (which OpenClaw needs), installs it if missing, downloads the latest OpenClaw release, and sets up the directory structure.

Depending on your VPS, this might take a minute or two. When it finishes, you should see a message saying OpenClaw was installed successfully.

You can verify the installation by checking the version:

openclaw --version

If you see a version number, you are good to go.

For a deeper look at the platform features and what you can do with it, take a look at the AgentVPS features page.

Basic OpenClaw Commands

Once OpenClaw is installed, a handful of commands is all you need to get going. Here are the essentials.

Check Status

openclaw status

This tells you whether OpenClaw is running, stopped, or encountering an error. Run this first whenever something feels off.

Start and Stop

openclaw start
openclaw stop

Starting launches the daemon. Stopping shuts it down cleanly. Your agent does not do anything special during startup yet, but once you configure it, every start brings it back online.

View Logs

openclaw logs

Logs are your best debugging friend. If an integration is not working or a skill is failing, the logs will tell you why.

Deploy Configuration

openclaw deploy

This applies your current configuration file. You will edit your config to add integrations and skills. After every edit, run openclaw deploy to make the changes take effect.

Configuring OpenClaw: Your First Setup

OpenClaw uses a YAML configuration file. It lives at ~/.openclaw/config.yaml by default. Here is what a minimal working config looks like for a Telegram bot:

agent:
  name: "MyAgent"
  model: "claude-sonnet-4-20250514"

platforms:
  telegram:
    bot_token: "YOUR_BOT_TOKEN_HERE"

The agent section names your agent and picks the LLM model it uses. The platforms section connects your agent to messaging apps.

You add one bot token per platform. Each platform has its own setup steps, which we cover in the next section.

Connecting Telegram, Discord, or Slack

OpenClaw supports multiple messaging platforms out of the box. You pick the ones you want.

Telegram

1. Open Telegram and search for @BotFather
2. Send /newbot and follow the prompts
3. BotFather gives you a bot token. It looks like 123456:ABC-DEF1234ghIkl-zyx57W2v1u123ew11
4. Put that token in your config.yaml under platforms.telegram.bot_token

After you run openclaw deploy, your bot will come online. Message it on Telegram. If it replies, you are connected.

Discord

1. Go to the Discord Developer Portal and create a new application
2. Go to the Bot section and create a bot
3. Copy the bot token
4. Enable the Message Content Intent (required for the bot to read messages)
5. Invite the bot to your server using the OAuth2 URL generator with bot and applications.commands scopes
6. Add the token to your config under platforms.discord.bot_token

Slack

1. Create a Slack app at api.slack.com
2. Add the Bot Token scope under OAuth & Permissions
3. Install the app to your workspace
4. Copy the Bot User OAuth Token (starts with xoxb-)
5. Add it to your config under platforms.slack.bot_token

Each platform has slightly different setup, but the pattern is always the same. Create a bot, get the token, add it to the config, deploy. That is it.

If you need a managed setup where all of this is handled for you, check out our guide on what is managed OpenClaw hosting.

Setting Up Skills

Skills are where OpenClaw gets its superpowers. A skill is essentially a set of instructions plus the tools your agent can use.

Out of the box, OpenClaw comes with built-in skills. These include:

• Web search -- lets your agent search the internet for current information
• Web fetch -- reads and summarizes web pages
• File system -- reads and writes files on the server
• Memory -- stores and recalls information across conversations

You enable skills in your config file:

skills:
  web_search:
    enabled: true
  web_fetch:
    enabled: true
  memory:
    enabled: true

You can also install community skills or write your own. A custom skill might check your email, pull calendar events, control smart home devices, or run custom scripts on your server.

To install a community skill, you typically clone a repository or add a package. Each skill comes with its own documentation. OpenClaw's skill system is one of its strongest features because you are not limited to what comes pre-installed. You build exactly the agent you want.

For a broader overview of what is possible, visit the AgentVPS about page.

OpenClaw Tips for Beginners

Here are some things I wish someone had told me when I started.

Start Simple

Do not try to connect every platform and install every skill at once. Start with one platform and one skill. Get that working. Then add more. The incremental approach saves you from debugging five things at once.

Keep Your Bot Tokens Secret

Your bot tokens are essentially passwords. If someone gets your Telegram bot token, they can control your bot. Keep your config.yaml file secure. Never commit it to a public GitHub repository.

Use Screen or Tmux

OpenClaw runs as a daemon, but during testing you might want to see its output in real time. Tools like screen or tmux let you run processes that stay alive even if you disconnect from SSH. Start one of these before launching OpenClaw during initial setup.

Check Logs First

Something not working? Run openclaw logs before anything else. Nine times out of ten, the error message in the logs tells you exactly what is wrong -- a missing token, a wrong port, a misconfigured skill.

Name Your Agent Something Fun

Your agent is going to be your digital companion. Give it a personality. Name it something you enjoy saying. You can change the name in the config file under agent.name.

Common Questions About OpenClaw

Do I need coding experience to use OpenClaw?

Not really. Basic familiarity with the command line helps, but the configuration is all YAML. Editing a text file and running a deploy command is about as technical as it gets for the basics. If you can edit a settings file, you can use OpenClaw. Writing custom skills does require some coding, but you can use community skills without writing a single line of code.

Which LLM should I use with OpenClaw?

OpenClaw supports multiple models including Claude, GPT, Gemini, and local models. For most users, Claude Sonnet is a great starting point. It balances capability, speed, and cost well. You can always switch models later by changing one line in your config file. Experiment and see which one fits your use case best.

Is OpenClaw free?

Yes, the OpenClaw framework itself is open source and free. You pay for your VPS hosting and you pay for API usage from whichever LLM provider you choose. The cost is generally quite low for personal use. Think pennies per day for a moderately active agent.

Can I run multiple agents on one VPS?

Yes. You can run separate OpenClaw instances with different config files, each connecting to different bots and platforms. Just make sure your VPS has enough memory and CPU to handle all of them. Each agent instance needs its own configuration directory and its own port.

What happens if my VPS goes down?

Your agent goes offline. That is one reason to choose reliable hosting. Most VPS providers offer uptime guarantees and automatic restart. When the VPS comes back up, you can set OpenClaw to start automatically. AgentVPS handles this with health monitoring and automatic restarts as part of its managed offering.

Your Next Steps

You have the basics. You know what OpenClaw is, how to set it up on a VPS, connect platforms, and enable skills. What comes next is the fun part. Customization.

Build your agent. Give it a personality. Connect the tools you use every day. Let it handle the repetitive stuff so you can focus on the things that matter.

OpenClaw is powerful because it is yours. It does not serve ads, it does not phone home, it does not sell your data. It just works for you, on your terms.

Ready to Deploy Your Own AI Agent?

You have seen how OpenClaw works and what it can do. The only thing left is to get your own instance running. AgentVPS makes it dead simple. No complex server setup, no wrestling with dependencies, no late-night debugging sessions.

We provide VPS hosting optimized for AI agent workloads. That means the right memory, the right CPU, and the right network configuration for your OpenClaw instance. You get root access, 24/7 uptime, and the peace of mind that your agent is always online.

Get started with AgentVPS today. Spin up your server in minutes. Install OpenClaw in seconds. Your AI agent is waiting.