How to Run Your Own Private AI on WhatsApp for $5 a Month: OpenClaw VPS Setup Guide

What You Are Building

By the end of this guide you will have OpenClaw running 24/7 on a cloud server, accessible from WhatsApp or Telegram, connected to an AI model of your choice. Your messages go to your phone. Your AI assistant responds, remembers context, and takes actions.

Total cost: approximately $5–$10/month for the VPS, plus AI API costs (typically $5–$20/month depending on usage).

Before You Start: What You Need

• A VPS (virtual private server) — we will use Hetzner or DigitalOcean
• An API key from an AI provider: Anthropic (Claude), OpenAI, or DeepSeek
• A messaging app account: WhatsApp, Telegram, or Discord
• Basic comfort with a terminal — you will run some commands, but nothing complex

If you have never used a terminal before, this guide will work — just follow each step carefully.

Step 1: Get a VPS

Recommended Providers

Hetzner (cheapest, European servers, excellent for privacy)

• CX22 plan: 4GB RAM, 2 vCPU, €4.35/month
• Good default choice for most people

DigitalOcean (beginner-friendly, has a 1-click OpenClaw droplet)

• Basic Droplet: 2GB RAM, 1 vCPU, $12/month
• Slightly pricier but easier UI and has native OpenClaw support

Vultr or Linode — similar pricing and specs to DigitalOcean, all work fine.

Minimum Specs

OpenClaw's documentation says 1GB RAM is enough. In practice, 4GB RAM and 2 vCPU is the minimum for smooth operation. With 1GB you will run into memory issues.

Creating Your Server

• Sign up at Hetzner or DigitalOcean
• Create a new server with Ubuntu 22.04 LTS as the operating system
• Choose the 4GB RAM plan
• Add your SSH key (or use password login — SSH key is more secure)
• Note the server's IP address once it is created

Step 2: Connect to Your Server

Open your terminal and connect via SSH:

ssh root@YOUR_SERVER_IP

Replace YOUR_SERVER_IP with the IP address from your VPS dashboard. If you are on Windows, use Windows Terminal or install PuTTY.

You should see a command prompt like root@ubuntu-server:~# — you are connected.

Step 3: Update Your Server

Always update before installing anything:

apt update && apt upgrade -y

This takes 1–2 minutes. Wait for it to complete.

Step 4: Install OpenClaw

Run the official install script:

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

This downloads and installs OpenClaw and all its dependencies (Node.js 22+, required packages). The installation takes 3–5 minutes.

When it completes you should see:

OpenClaw installed successfully.
Run 'openclaw setup' to configure your assistant.

Step 5: Configure OpenClaw

Run the setup wizard:

openclaw setup

This walks you through the configuration interactively.

Choose Your AI Model

You will be asked which AI model to use. Your options:

Claude (Anthropic) — Best reasoning, excellent at writing and analysis, most reliable for complex tasks. API key from console.anthropic.com. Cost: approximately $0.003 per 1,000 tokens.

GPT-4o (OpenAI) — Strong all-rounder, wide tool support, familiar to most users. API key from platform.openai.com. Cost: approximately $0.005 per 1,000 tokens.

DeepSeek — Cheapest option, surprisingly capable, good for coding tasks. API key from platform.deepseek.com. Cost: approximately $0.0001 per 1,000 tokens — roughly 30x cheaper than GPT-4.

For most users: start with DeepSeek to keep costs low while you learn the system. Switch to Claude or GPT-4 once you know what you want.

Enter your chosen API key when prompted.

Choose Your Messaging Platform

Telegram — Easiest to set up. No QR code needed. Create a Telegram bot via @BotFather, get the bot token, paste it into the setup. Recommended for beginners.

WhatsApp — More complex. A QR code appears in your terminal. You scan it from WhatsApp on your phone (Settings → Linked Devices → Link a Device). Uses WhatsApp Web protocol — works but is technically unsupported by Meta and can occasionally disconnect.

Discord — Create a Discord application, get a bot token, invite the bot to your server. Good if you already use Discord heavily.

For beginners: use Telegram. It is the most stable and easiest to configure.

Set Your Name and Timezone

OpenClaw will ask for your name (for personalised greetings) and timezone (for scheduling and reminders). Enter these accurately — timezone especially matters for morning briefings and time-based reminders.

Step 6: Connect Your Messaging App

Telegram Setup

• Open Telegram and search for @BotFather
• Send /newbot and follow the prompts to create a bot
• Copy the bot token BotFather gives you
• Paste it into the OpenClaw setup when asked
• Start a chat with your new bot — search for the bot name you chose

WhatsApp Setup

• During setup, a QR code will appear in your terminal
• Open WhatsApp on your phone
• Go to Settings → Linked Devices → Link a Device
• Scan the QR code
• Wait for the connection to confirm

The WhatsApp session can occasionally disconnect (especially after phone restarts or WhatsApp updates). If it disconnects, run openclaw reconnect on your server.

Step 7: Start OpenClaw

openclaw start

To run it as a background service that persists after you close your terminal:

openclaw service install
openclaw service start

Now OpenClaw runs automatically, even if the server restarts.

Step 8: Test It

Send your bot a message on Telegram (or WhatsApp):

Hello, what can you do?

You should get a response within a few seconds listing its capabilities.

Try a simple task:

What is the weather in New Delhi today?

Or:

Remind me at 9am tomorrow that I have a team meeting

If both work, your setup is complete.

Useful Commands

openclaw status        # Check if OpenClaw is running
openclaw logs          # See recent activity and errors
openclaw restart       # Restart the service
openclaw update        # Update to the latest version
openclaw reconnect     # Reconnect messaging platform (if disconnected)

Security: What You Must Do

OpenClaw has access to your messaging apps, potentially your email, calendar, and terminal. This is powerful — and requires care.

Use a dedicated phone number for WhatsApp. Do not link OpenClaw to your primary WhatsApp account. If something goes wrong, you do not want your main contacts affected. A cheap SIM for a secondary number costs very little.

Create a dedicated Telegram bot. Never link OpenClaw to your personal Telegram account — always use a bot account (which is what @BotFather creates).

Limit terminal access. In openclaw.config.json, set terminal.enabled: false unless you specifically need it. Terminal access means OpenClaw can run commands on your server — useful but a meaningful security surface.

Keep your server updated. Run apt update && apt upgrade -y monthly. OpenClaw only works safely on a maintained server.

Use a firewall. Block all ports except 22 (SSH) and whatever port OpenClaw uses for webhooks:

ufw allow 22
ufw allow 3000
ufw enable

Back up your config. Your OpenClaw configuration and memory files are in ~/.openclaw/. Copy this directory somewhere safe periodically.

Monthly Cost Estimate

• VPS (Hetzner CX22): ~$5/month
• AI API (DeepSeek, moderate usage): ~$2–$5/month
• AI API (Claude, moderate usage): ~$10–$20/month
• Total: $7–$25/month depending on usage and model choice

This is significantly cheaper than any commercial AI assistant subscription that offers comparable capability.

Troubleshooting Common Issues

OpenClaw installed but not responding to messages

Check if the service is running: openclaw status. If it shows stopped, run openclaw service start.

WhatsApp keeps disconnecting

This is a known issue with the WhatsApp Web protocol. Either accept periodic reconnections (openclaw reconnect) or switch to Telegram, which is more stable.

Out of memory errors

Your VPS has insufficient RAM. Upgrade to the 4GB plan. 1–2GB is not enough for smooth operation.

API rate limit errors

You are sending too many messages too quickly. DeepSeek has generous rate limits. OpenAI and Anthropic have stricter ones on lower-tier plans. Check your API provider's dashboard.

Messages delivering but responses are slow (10+ seconds)

Normal for complex tasks. For simple queries, responses should be 2–4 seconds. If simple queries are slow, check your server's CPU with top.

What to Try First

Once your setup is working, here are the most useful things to test:

Morning briefing: "Set up a daily morning briefing at 8am with today's weather, my calendar events, and top tech news headlines."

Email summary: Connect your Gmail or Outlook in the OpenClaw settings and try: "Summarise my unread emails from the last 24 hours."

Research: "Research the top 5 competitors to Notion and give me a comparison table."

Reminder with context: "Remind me on Friday morning to follow up with the client about the proposal I sent on Monday."

Code help: "Review the error in the logs at ~/.pm2/logs/app-error.log and suggest a fix."

The depth of what OpenClaw can do becomes clear quickly once it is running. Start simple, add integrations gradually, and let it build context about your routines and preferences over time.