Most OpenClaw tutorials are written by developers, for developers. They assume you're comfortable in a terminal and know what environment variables are.
This one doesn't. I'm not a developer either.
This tutorial uses the official openclaw onboard wizard — the recommended starting path for new users — and walks through each step in plain English. By the end, you'll have a running agent connected to Telegram with at least one skill installed.
What You'll Build
A personal AI agent that:
- Runs on a server (yours, not OpenClaw's)
- Responds to messages via Telegram
- Has at least one skill (we'll use a basic utility skill)
- Runs persistently — available when you're not at your computer
What You Actually Need
Node.js 22 or newer. OpenClaw is Node.js-based. You need it installed on whatever computer will run the agent. On a VPS, you'll install it during setup. On a Mac, install it from nodejs.org.
A server OR your local machine. For testing, you can run OpenClaw on your own computer. For production use, you need a server — a computer that runs 24 hours a day. Hetzner CX22 (~€5/month) is the most common choice. The rest of this tutorial assumes you're using a Hetzner VPS.
An API key. OpenClaw uses your Anthropic or OpenAI account. You'll need an account with billing enabled. Don't start with the most expensive model — the default settings are fine.
A Telegram account. This tutorial connects your agent to Telegram. You'll need a Telegram account and a Telegram bot token. We'll create the bot in step 3.
Time. Realistically: 2–4 hours for a first-time setup. Not 20 minutes. Not a weekend. But not overwhelming if you follow these steps.
Step 1: Prepare Your Server
If you're using Hetzner:
- Create a Hetzner account at hetzner.com
- Click "Add server"
- Choose: Ubuntu 22.04, CX22 (2 vCPU, 4GB RAM), your region
- Under "SSH keys," add your public key (on Mac, it's in
~/.ssh/id_rsa.pub— create one if you don't have it) - Name your server and click "Create & Buy Now"
Note the IP address shown after creation. You'll need it.
Connect to your server:
ssh root@YOUR_SERVER_IP
Install Node.js 22:
curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
apt-get install -y nodejs
node --version # should show v22.x.x
Step 2: Install OpenClaw
npm install -g openclaw
Verify the installation:
openclaw --version
If you see a version number, you're ready for the next step.
Step 3: Create Your Telegram Bot
Before running the wizard, create a Telegram bot token. You'll need this during the onboarding.
- Open Telegram and search for BotFather (the official bot, with a blue checkmark)
- Start a conversation with BotFather
- Send:
/newbot - Choose a name for your bot (e.g., "My Personal Assistant")
- Choose a username — must end in "bot" (e.g., "mypersonalassistant_bot")
- BotFather sends you a token: a long string like
1234567890:ABCdef...
Copy that token. You'll paste it during the wizard.
Step 4: Run the Onboard Wizard
This is the recommended path. The wizard asks you questions and handles the configuration:
openclaw onboard --install-daemon
The --install-daemon flag sets up OpenClaw to run as a background service — so it keeps running after you close your terminal and restart when your server reboots. Use it.
The wizard will ask you:
Your API key. Paste your Anthropic or OpenAI key here.
Your primary channel. Choose Telegram (or whichever channel you're setting up first).
Your Telegram bot token. Paste the token from BotFather.
Your model preference. Accept the default for now. You can change it later.
Your workspace directory. Accept the default (~/.openclaw/workspace/). This is where your agent stores context.
When the wizard finishes, it starts your agent as a daemon. You should see a confirmation message.
Step 5: Test the Connection
Open Telegram. Search for the bot username you created in step 3. Start a conversation.
Send: Hello
Your agent should respond. If it does, the connection is working.
If it doesn't respond after 30 seconds:
- Check that the daemon is running:
openclaw gateway status - Check the logs:
openclaw logs --follow - Verify your Telegram token is correct:
openclaw config
Step 6: Install Your First Skill
Skills are what give your agent capabilities beyond conversation. They're installed from ClawHub — the OpenClaw skill registry at clawhub.ai.
Telegram alone has 145,000+ installs on ClawHub, making it the most popular skill. But you've already set up Telegram through the wizard, so let's install something that adds real capability.
A good first skill is AgentMail — it connects your email so your agent can read and summarize messages.
clawhub install agentmail
The skill will prompt you for your email credentials. Follow the skill's own setup instructions.
After installation, test it in Telegram:
What's in my inbox today?
If the skill is working, your agent will check your email and respond with a summary.
Step 7: Run Your Agent as a Gateway (Optional)
If you want your agent to be accessible via webhook — useful for some integrations — you can run it in gateway mode:
openclaw gateway
This starts a web server that receives incoming requests. For most users, this isn't needed initially. The daemon mode from step 4 is sufficient.
Common Setup Problems
Agent doesn't respond after setup.
Check openclaw gateway status. If it shows "stopped," run openclaw gateway run. If it shows running but still no response, check openclaw logs --follow for error messages.
Telegram shows "bot is not responding."
This means Telegram is reaching your bot, but your agent isn't processing the message. Usually means the daemon crashed. Check logs, then restart: openclaw gateway restart.
"Invalid API key" error. Verify your API key has billing enabled. Even valid keys fail if the account has no payment method. Double-check that you copied the key exactly — no extra spaces.
Skills not loading after install.
Run openclaw gateway restart to force the agent to pick up new skills. If a skill still doesn't work, check its ClawHub page for version compatibility notes.
High API costs immediately. You haven't configured model routing. Without it, all requests use your configured default — often a premium model. OpenClaw docs cover model routing under "Concepts." Set cheaper models for simpler tasks.
What Comes Next
You have a running agent. A few things worth doing in the next week:
Configure security. The wizard creates a basic setup, but you should review the firewall and authentication settings before considering this production-ready. Read the security guide →
Add more skills. Browse clawhub.ai. Start with 2–3 well-maintained skills (check the last-updated date). Test each before adding the next. Many skills conflict.
Explore Telegram commands. Try /help in your Telegram chat to see what commands your agent understands. The more you use it, the more it learns your patterns.
Set up monitoring. Your agent will eventually crash or fall behind on updates. Set up a simple uptime monitor (UptimeRobot has a free tier) to alert you if your agent stops responding.
If you'd rather skip all of this: Volos handles the setup, security, skill selection, and ongoing maintenance through conversation. See the pricing →
Learn what skills are available and which to install first →
