Your new WHC VPS comes with OpenClaw pre-installed — a self-hosted AI assistant that lives on your server and answers you through the messaging apps you already use. No separate AI app to download, no command line required for typical setups, and your data stays on your VPS except when sent to your chosen AI provider.
This guide will take you from a freshly provisioned VPS to a working bot on Telegram that can summarize any webpage you send it. Set aside about fifteen minutes and a cup of coffee, and by the end you'll have a real, useful AI assistant in your pocket.
What OpenClaw does
Think of OpenClaw as a bridge. On one side: the messaging apps you already use — Telegram, Slack, Discord, WhatsApp, Teams, and others. On the other side: the AI model of your choice — Anthropic's Claude, OpenAI's GPT, Google's Gemini, or even a model running locally. OpenClaw sits in the middle, sending your messages to the AI and returning the responses back to you.
The important part: all of this runs on your VPS, in its own isolated LXD container. Your conversations, your configuration, and your API keys stay with you.
Before you start
You'll need three things:
An active WHC VPS with OpenClaw included. If you're reading this, you probably already have one.
An AI provider account. This guide uses Anthropic because it's the simplest to set up and works beautifully with OpenClaw. You can swap in another provider later. Sign up at console.anthropic.com and add a small amount of credit — $5 is usually enough for light testing and initial use.
A Telegram account on your phone. If you don't have one, install Telegram from the App Store or Google Play. It's free.
That's it. No SSH clients, no config files, no package managers.
Procedure
Step 1 — Log in to OpenClaw
Navigate to your VPS product.
Click the Open OpenClaw button. We handle the login for you through single sign-on, so you won't need a separate password — clicking the button drops you straight into the OpenClaw dashboard.
The first time you land, you'll see the OpenClaw Overview page: a summary of your gateway's health, connected channels, and active agents. There won't be much happening yet — that's about to change.
Step 2 — Connect your AI provider
Note: OpenClaw doesn't ship with an AI model of its own. You need to tell it which one to use and give it a key to access it.
In the OpenClaw dashboard, go to the Models page from the left sidebar. You'll see a list of supported providers: Anthropic, OpenAI, Google Gemini, Groq, OpenRouter, and a few more. Click Anthropic.
You'll be asked for an API key. Here's how to get one:
Log in to console.anthropic.com.
Click your profile icon in the top-right, then API Keys.
Click Create Key.
Give it a name (example: "My OpenClaw VPS").
Copy the key it shows you. It'll look like a long string starting with sk-ant-.
Go back in OpenClaw.
Paste the key into the API key field.
Click Save.
OpenClaw will test the connection and confirm it works. You'll also be asked to pick a default model — Claude Sonnet is a sensible starting point: fast, capable, and affordable.
If the connection fails, double-check that your API key was copied correctly and that your provider account has available credit.
A word on other providers. If you'd rather use OpenAI, Gemini, or a local model via Ollama, the process is nearly identical — pick the provider on the Models page and follow the prompts. The rest of this guide works the same regardless of which provider you chose.
Step 3 — Set up your Telegram bot
Telegram has a clever little system for creating bots: you talk to a bot that makes bots. It's called BotFather, and it's the fastest way to get a Telegram bot up and running.
On your phone:
Open Telegram.
Search for @BotFather. Make sure it's the verified one — it has a blue checkmark.
Start a chat with BotFather.
Send the command /newbot.
BotFather will ask for a name for your bot. This is the display name that shows up in chats. Something friendly works well — "My Assistant," "Gord," whatever feels right.
Next, BotFather will ask for a username. This has to be unique across all of Telegram and must end in bot (for example, my_assistant_bot). If the first name you try is taken, just try another.
Once you've settled on a username, BotFather will send you a message containing your bot token. It looks something like 7891234567:AAHk8xxxxxxxxxxxxxxxxxxxxxx. Copy it and keep it safe — this token is essentially the password to your bot.
Back in OpenClaw:
Go to the Channels page from the left sidebar.
Click Telegram.
Click Add Channel.
Paste your bot token into the field provided.
Click Save.
Within a few seconds, OpenClaw will connect to Telegram and the channel status will switch to Connected. Your bot is live.
Note: Before using your bot, make sure to secure it in the next step.
Step 4 — Lock it down before you use it
IMPORTANT:
Don't skip this step. A bot token on the open internet is a bot anyone can talk to — and since your bot is connected to an AI model that you're paying for, anyone talking to it is spending your money. Worse, they're talking to an assistant that can see whatever tools you've connected to it.
OpenClaw makes it easy to lock this down with an allowlist (shown as “Allowed Users” in the UI): a list of Telegram usernames or user IDs that are allowed to message the bot. Anyone else gets ignored.
Still on the Channels page, click into your Telegram channel's settings.
Find the Allowed Users (or allowFrom) field.
Add your own Telegram username — the one that starts with @. For example, @jane_doe.
Click Save.
Note: If you want to let a few family members or colleagues in too, add their usernames to the same list. Everyone else — spammers, strangers, bots that crawl BotFather output looking for exposed tokens — will be politely ignored.
If you ever think your token has leaked, go back to BotFather on Telegram, send /revoke, pick your bot, and it'll issue a new token. Paste the new token into OpenClaw's channel settings and you're safe again.
Note: If your bot isn’t responding later, this is the first place to check.
Step 5 — Say hello
Now, the fun part, testing your bot:
Open Telegram on your phone.
Search for your bot by username (the one ending in bot).
Tap Start or send any message (something like "hello," "are you there?").
Within a few seconds, your bot should reply. You've just sent a message from your phone, through Telegram, to your VPS in Canada, through OpenClaw, out to Anthropic's API, and back again. All in about the time it takes to blink.
Congratulations — you have an AI assistant.
Note: If your bot doesn’t respond, confirm the channel shows “Connected” in OpenClaw and that your Telegram username is included in the Allowed Users list.
Step 6 — Make it do something useful: summarize a webpage
A chatbot that just answers trivia is neat. A chatbot that can go read things on the web for you is actually useful.
OpenClaw ships with a set of built-in skills — self-contained capabilities that give your AI new powers. One of them is web browsing. To confirm it's enabled:
In the OpenClaw dashboard, open the Skills page.
Find the Web skill (sometimes labeled "browser" or "web-fetch" depending on your version).
Make sure the toggle is On. If it isn't, flip it and save.
Now, back in Telegram, send your bot a message like:
Summarize this article in three bullet points: https://whc.ca/blog/is-your-website-hosting-slowing-down-your-business/
Your bot will fetch the page, read it, and send back a clean summary. Try it with a news article, a long product review, a blog post you've been meaning to get to — anything on the open web. If the summary fails, make sure the Web skill is enabled and that the page you’re sending is publicly accessible.
You've officially gone from zero to a real AI assistant that does a real job.
Keeping things secure and affordable
Now that everything works, a few quick habits will save you grief down the line.
Protect your API key. Your Anthropic key is stored inside your OpenClaw container on your VPS — never in Telegram, never in a public place. That said: don't paste it into emails, Slack messages, or support tickets. If you ever need to share your OpenClaw setup with someone, share screenshots with the key blurred out. If you suspect a key has leaked, revoke it immediately at console.anthropic.com and generate a new one.
Set a spending limit. Anthropic and most other AI providers let you cap your monthly spend. Log in to your provider's dashboard and set a hard limit — $10 or $20 a month is plenty for casual use and means you'll never get a surprise bill. This is the single best thing you can do to sleep well at night.
Know what your bot can touch. The OpenClaw container on your VPS runs in its own isolated LXD container — with no default access to your other services. That said, as you connect more skills and integrations (Google Drive, calendars, your own code), you expand what the AI can see and do. Add them one at a time, and always ask yourself: "Am I comfortable with this assistant having access to this?" When in doubt, don't connect it.
Keep your Allowed Users (allowlist) tight. Only add people you trust. A bot with access to an AI model is effectively a bot with access to your wallet.
What to try next
You now have a working, secure, useful AI assistant. Some natural next steps:
Connect more channels. The same process works for Slack, Discord, WhatsApp, and Microsoft Teams. Add them from the Channels page.
Enable more skills. OpenClaw has skills for reading files, managing calendars, sending emails, and dozens of other things. Browse the Skills page and turn on what interests you.
Explore agents. An "agent" in OpenClaw is a customized personality and skill set — you can have one agent for work, another for personal tasks, each with different tools and a different tone.
And if you get stuck, our support team is here 24/7 — open a ticket from your Client Area or reach out through live chat.
Welcome to your new assistant. Enjoy.