Table of Contents
From Zero to a 24/7 Personal AI Assistant
The setup conversation for Clawbot starts with the same question every time: “Does it really work?”
The answer is yes. But more importantly, it works differently than every other AI tool you’ve tried.
Unlike web-based AI assistants that disappear when you close the tab, a personal AI assistant like Clawbot stays online. It remembers what you said last week. It can reach out to you first. And it can control your computer the way a trusted colleague would.
This guide takes you from zero to running your own personal AI assistant in about 30 minutes. Whether you’re on macOS, Linux, or Windows, we’ll walk through every step—from installing the runtime to sending your first message.
By the end, you’ll understand why Mac Minis are selling out and why developers are replacing their Zapier subscriptions with a personal AI assistant that costs less per month.
What Makes Clawbot Different From Other AI Assistants
Before diving into installation, let’s clarify what you’re actually setting up.
Most AI assistants operate in a confined space. They’re good at answering questions. They’re terrible at doing actual work. They forget context. They require constant re-prompting. And they live on someone else’s servers.
A personal AI assistant like Clawbot inverts that model entirely.
Clawdbot isn’t just a chatbot — it automates tasks, remembers your preferences, and integrates with WhatsApp, Discord, Slack, and more. Learn how it became the viral AI product of 2026 here.
The Four Pillars of a True Personal AI Assistant
Persistence: The assistant stays online continuously. You can assign tasks at 2 AM and check the results when you wake up. Most web-based AI assistants can’t do this.
Memory That Compounds: Every conversation is summarized. Important facts are extracted. Preferences are stored. A week later, you mention something tangentially related, and the AI already knows the context. This is the difference between a tool and a true assistant.
Proactive Communication: The assistant doesn’t just wait for your input. It can send you morning briefings. It can alert you to calendar conflicts. It can flag important emails before you see them. This is what “24/7” actually means.
Real Computer Access: The assistant can browse, click, type, and execute commands. It can fill forms. It can send emails under your account. It can integrate with your actual workflow, not a sandbox version of it.
Most commercial AI assistants miss at least three of these four pillars. A personal AI assistant like Clawbot has all of them.
What You Actually Need to Get Started
Let’s be direct about requirements. The installation process sounds technical, but most of the complexity is handled automatically.
Hardware: Surprisingly Modest
The surprise when people first read Clawbot’s specs is that they look fake:
- 1GB RAM (2GB recommended for comfort)
- 500MB of disk space
- Single CPU core
- Any computer from the last decade
This isn’t a mistake. A personal AI assistant doesn’t need much horsepower. It needs persistence.
Where to run it:
- Old laptop from 2015? Works.
- Raspberry Pi 4? Works.
- Mac Mini? Works great—and is currently the popular choice.
- VPS at $5/month? Also works—and actually preferred by many for always-on reliability.
The Mac Mini trend exists because of convenience and iMessage support, not because it’s technically necessary.
Software: The Minimum Viable Stack
You need three things:
- Node.js 22 or newer — The JavaScript runtime that powers the entire system
- Git — For version control (technically optional, but recommended)
- A terminal — Command line access to your machine
That’s it. No Docker. No complex container orchestration. No virtual environments. The onboarding wizard handles everything else.
The Real Cost: AI Model Access
Here’s where the actual expense lives. You need access to an AI model—the brain behind your personal AI assistant.
Three routes:
| Option | Cost | Best For |
|---|---|---|
| Anthropic API | $5-$25 per million tokens (pay-as-you-go) | Light to medium use ($20-80/month typical) |
| Claude Pro | $20/month | Casual daily use with built-in limits |
| Claude Max | $100/month | Heavy users burning millions of tokens |
| OpenAI API | Similar to Anthropic API | Alternative if you prefer GPT-4 |
| Local Models | Free | Privacy-first, but lower quality |
Most people starting out pick Claude Pro ($20/month) or Anthropic API with a set budget. Heavy users switch to Claude Max after a week when they realize how much work the AI can compress.
Phase 1: Install Node.js (Your Runtime Engine)
Clawbot runs on Node.js, the same technology that powers Netflix, Slack, and thousands of enterprise applications. You need version 22 or newer.
macOS Installation
If you use Homebrew (most macOS users do):
brew install node@22
node --version
npm --version
You should see output like v22.x.x.
If you don’t have Homebrew, get it at brew.sh, or use nvm instead:
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.0/install.sh | bash
source ~/.zshrc
nvm install 22
nvm use 22
Linux Installation (Ubuntu/Debian)
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt-get install -y nodejs
node --version
Windows Installation (Critical: Use WSL2)
Native Windows support for Clawbot is limited and problematic. Use Windows Subsystem for Linux 2 instead:
wsl --install -d Ubuntu
After your system restarts, open Ubuntu from your Start menu and follow the Linux installation instructions above.
If you absolutely must use native Windows, download from nodejs.org, but expect more issues and less community support. WSL2 is the tested path.
Phase 2: Install Clawbot Itself
With Node.js running, installing Clawbot is genuinely one command.
The Easy Route (Recommended)
curl -fsSL https://clawd.bot/install.sh | bash
clawdbot --version
This one-liner downloads the installer, runs it, and sets everything up. You should see version output like 2026.1.24.
The Manual Route (If You Prefer)
npm install -g clawdbot@latest
Or if you use pnpm:
pnpm add -g clawdbot@latest
Either way, verify it worked:
clawdbot --version
Phase 3: The Onboarding Wizard (Where Everything Comes Together)
This is the magic step. One command guides you through everything.
clawdbot onboard --install-daemon
The --install-daemon flag means Clawbot keeps running in the background, even when you close the terminal. This is essential for a 24/7 personal AI assistant.
What the Wizard Asks (And What to Choose)
Step 1: Authentication Method
You’ll see options:
? Authentication method:
❯ Anthropic API Key (recommended for starters)
OpenAI API Key
OAuth (Claude Pro/Max subscription)
Local Models
For most people: choose Anthropic API Key. Get your key at console.anthropic.com, then paste it when prompted.
If you already have Claude Pro or Max, choose OAuth and let it authenticate through your browser.
Step 2: Model Selection
? Default model:
❯ anthropic/claude-opus-4-5 (most capable)
anthropic/claude-sonnet-4-5 (balanced)
openai/gpt-4-turbo
openai/gpt-4o
Honest advice: Opus 4.5 is overkill for most tasks. Choose Sonnet 4.5 for 80% of use cases. It costs less and responds faster. You can always switch later.
Step 3: Channel Selection
? Which channels should your AI assistant use?
◉ WhatsApp
◯ Telegram
◯ Discord
◯ Slack
◯ Signal
Pick at least one. WhatsApp is most popular. You can add others later. This is just selecting which messaging apps will connect to your personal AI assistant.
Step 4: Workspace Directory
Default is fine. This is where your AI assistant stores memory, preferences, and files.
? Workspace directory: ~/clawd
Step 5: Gateway Configuration
For home hardware:
? Gateway type:
❯ Local Gateway (runs on your machine)
Remote Gateway (cloud-based)
Choose Local Gateway. This keeps everything on your hardware.
Step 6: Port Configuration
Default port 18789 is fine unless you have conflicts:
? Gateway port: 18789
Phase 4: Connect Your First Channel (WhatsApp)
WhatsApp is the most intuitive entry point for controlling your personal AI assistant from your phone.
The QR Code Setup
Run this command:
clawdbot channels login
A QR code appears in your terminal. You have 60 seconds to scan it.
On your phone:
- Open WhatsApp
- Go to Settings → Linked Devices → Link a Device
- Point your phone at the terminal QR code and scan it
- Done
That’s it. WhatsApp is now connected to your personal AI assistant.
The Phone Number Question
Best practice: Use a separate phone number for your AI assistant. Get an eSIM or use an old phone. This keeps your personal and AI contact lists separate.
Acceptable: Use WhatsApp Business with a different number on the same device.
Not recommended but possible: Use your main WhatsApp number, but be careful. Every incoming message becomes input to your AI assistant.
The Security Layer: Pairing System
By default, unknown numbers can’t just start commanding your AI assistant.
When an unknown contact messages it, they receive a time-limited pairing code. The message is ignored until you approve them:
clawdbot pairing approve whatsapp ABC123XY
Check pending requests:
clawdbot pairing list whatsapp
This is a simple but effective security feature that prevents random people from accidentally using your personal AI assistant.
Phase 5: Verify Everything Works
Test that your personal AI assistant is actually running:
clawdbot gateway status
You should see:
✅ Gateway listening on ws://127.0.0.1:18789
✅ Model: anthropic/claude-sonnet-4-5
✅ WhatsApp: connected
Send a test message from your phone to your linked number:
You: Hello! Can you hear me?
AI Assistant: Hello! I'm your personal AI assistant.
I can help with questions, automate tasks, manage your calendar,
and much more. What would you like help with?
If you get a response within a few seconds, your personal AI assistant is working.
Alternative Channels: Beyond WhatsApp
Not everyone wants WhatsApp. Here’s how to add other channels to your personal AI assistant.
Telegram Setup
Telegram is often simpler because it has proper bot infrastructure.
- Message
@BotFatheron Telegram - Send
/newbotand follow the prompts - Copy your bot token
- During setup, select Telegram and paste the token
Or configure manually:
{
"channels": {
"telegram": {
"botToken": "123456:ABCDEFGH..."
}
}
}
Discord Setup
- Go to discord.com/developers/applications
- Create a new application
- Go to Bot → Reset Token and copy it
- Enable MESSAGE CONTENT INTENT (required)
- Paste the token during setup
Or manually:
{
"channels": {
"discord": {
"token": "your-discord-token"
}
}
}
Other Supported Channels
Clawbot also supports Slack, Signal, iMessage (macOS only), Microsoft Teams, Google Chat, Matrix, and others. Check the documentation for specific instructions.
The pattern is identical: authenticate with the service, add the credentials, and your personal AI assistant can operate through that channel.
Running Your Personal AI Assistant 24/7
For your personal AI assistant to send morning briefings, monitor emails, or execute scheduled tasks, it needs to run continuously.
Option 1: Dedicated Hardware (Mac Mini Approach)
A base Mac Mini ($500-600) has become the standard. It’s silent, uses minimal power, and iMessage support is a bonus.
Configure it:
- Disable sleep in System Preferences
- Enable automatic login
- Keep it plugged in
- The daemon (–install-daemon flag) ensures Clawdbot restarts on reboot
Some power users run 10+ Mac Minis for different personal AI assistant instances.
Option 2: VPS Hosting ($5/Month)
A virtual private server from Hetzner, DigitalOcean, or Linode is more reliable for 24/7 uptime.
Cheapest tier:
- 2GB RAM: $4-5/month
- Always on
- No electricity costs
- No home network dependency
After creating your VPS, SSH in and run the same installation steps. Your personal AI assistant runs in a data center instead of your home.
Option 3: Old Hardware You Already Own
That 2015 MacBook? A Raspberry Pi 4? A Windows laptop? All work fine.
Just set it to:
- Disable sleep mode
- Auto-login on boot
- Keep it plugged in
You lose 24/7 uptime when you shut it down, but gain zero hardware cost.
Essential Commands for Your Personal AI Assistant
Here are the commands you’ll use regularly:
# Check status
clawdbot gateway status # Is it running?
clawdbot doctor # Full diagnostics
clawdbot logs --follow # Watch real-time logs
# Channel management
clawdbot channels status # List connected channels
clawdbot channels login # Add WhatsApp
clawdbot channels logout # Remove a channel
# Pairing (security)
clawdbot pairing list whatsapp # See pending approvals
clawdbot pairing approve whatsapp CODE # Approve someone
# Updates
clawdbot update --channel stable # Update to latest stable
clawdbot update --channel beta # Try beta features
# Configuration
clawdbot configure # Re-run setup wizard
clawdbot configure --section web # Configure web search
Understanding Your Configuration Files
After setup, Clawbot creates files in ~/.clawdbot/:
~/.clawdbot/
├── clawdbot.json # Main configuration
├── credentials/
│ ├── oauth.json # OAuth tokens
│ └── whatsapp/
│ └── creds.json # WhatsApp session
└── agents/
└── main/
└── auth-profiles.json # API keys
Your personal AI assistant stores memory and preferences in ~/clawd/:
~/clawd/
├── SOUL.md # Your AI assistant's personality
├── MEMORY.md # Persistent memory (editable!)
├── skills/ # Extended capabilities
└── conversations/ # Chat history
You can edit SOUL.md to change how your personal AI assistant behaves and responds. You can also read MEMORY.md to see what it remembers about you.
Extending Your Personal AI Assistant With Skills
Skills are add-ons that extend what your personal AI assistant can do.
Popular skills include:
- Web Research: Summarize articles, pull data from websites
- Email Automation: Draft responses, categorize messages
- Calendar Integration: Check schedules, find meeting conflicts
- Code Review: Analyze pull requests, suggest improvements
- Browser Automation: Fill forms, click links, scrape data
- Smart Home: Control lights, thermostats, security systems
During onboarding, the wizard shows eligible skills:
│ Skills available: 10
│ Missing requirements: 39
│ Blocked by allowlist: 0
Install skills during setup or add them later through the configuration wizard.
Troubleshooting Common Issues
Gateway Won’t Start
clawdbot logs --follow
Look for error messages. Common causes:
- Missing Anthropic API key
- Port 18789 already in use on your machine
- Permission issues with your workspace directory
Fix:
clawdbot gateway restart
WhatsApp QR Code Keeps Refreshing
QR codes expire in 60 seconds. Scan quickly. If it keeps failing:
clawdbot channels logout
clawdbot channels login
Then try again.
“clawdbot is not recognized” (Windows)
Your npm global packages aren’t on your system PATH.
Find where npm installs globally:
npm config get prefix
Add that path + \bin to your Windows environment variables, then restart your terminal.
AI Assistant Responds Slowly
Try switching from Opus to Sonnet:
{
"agents": {
"defaults": {
"model": "anthropic/claude-sonnet-4-5"
}
}
}
Restart the gateway. Sonnet is faster and costs less.
Memory Isn’t Persisting
Check the MEMORY.md file in your workspace:
cat ~/clawd/MEMORY.md
If it’s empty, your personal AI assistant hasn’t been trained with memory extractions yet, or there are permission issues with the file. Give it a few conversations.
Security: The Trade-Offs of a Powerful Personal AI Assistant
Clawbot is powerful precisely because it’s unrestricted. This comes with real risks.
Core Principle: It’s Like Handing Someone Your Keys
Your personal AI assistant has shell access to your computer. It can read files. It can send emails. It can access saved passwords in your browser.
With great capability comes real responsibility.
Practical Security Measures
1. Don’t Install on Your Primary Machine First
Run it in a VM or on dedicated hardware while you’re learning.
2. Use Pairing Mode
Keep the pairing system on. Unknown contacts must be approved before they can use your personal AI assistant.
{
"channels": {
"whatsapp": {
"dmPolicy": "pairing"
}
}
}
3. Separate Credentials
Don’t save payment methods in the browser profile your personal AI assistant uses. Create a separate profile for automation.
4. Enable Allowlists for Sensitive Tools
{
"agents": {
"main": {
"skills": {
"allowlist": ["web_search", "email", "calendar"]
}
}
}
}
5. Regular Backups
Back up ~/.clawdbot/ and your workspace before major updates.
Real-World Use Cases for Your Personal AI Assistant
Here’s what people are actually building with a personal AI assistant like Clawbot:
Morning Briefings: Calendar, important emails, relevant news—delivered at 8 AM without asking.
Research Workflows: Summarize 10 articles, extract key data, compile into a report.
Email Triage: Unsubscribe from newsletters, categorize messages, draft responses to common questions.
Scheduling: Find meeting times across multiple calendars, send invites automatically.
Code Assistance: Review pull requests, write documentation, debug issues.
Memory Vault: One user transcribed 1,000+ voice messages and made them searchable through their personal AI assistant.
Flight Check-In: Automatically check in when the window opens, send confirmation.
Automation Chains: If a task involves multiple steps and decision-making, your personal AI assistant can handle it.
The common thread: if a human can do it on a computer, you can probably describe it to your personal AI assistant and it will figure out how.
Cost Breakdown: What Does Your Personal AI Assistant Really Cost?
Let’s be transparent:
| Component | Cost |
|---|---|
| Hardware | $0 (existing computer) to $600 (Mac Mini) |
| VPS alternative | $5-15/month (more reliable for 24/7) |
| AI Model (the real expense) | $20-200/month depending on usage |
| Anthropic API | $20-80/month (typical usage) |
| Claude Pro | $20/month (built-in limits) |
| Claude Max | $100/month (nearly unlimited) |
One early adopter burned through 180 million tokens in a week of heavy experimentation. Light users spend $15-20/month.
Bottom line: Hardware is optional. The AI model is the actual cost. Start with Claude Pro ($20/month) and upgrade if needed.
Updating Your Personal AI Assistant
The project moves fast—updates happen almost daily.
Check your version:
clawdbot --version
Update to stable:
clawdbot update --channel stable
Try beta features:
clawdbot update --channel beta
After updating, run diagnostics:
clawdbot doctor
Getting Help: The Community Is Incredibly Active
If you get stuck:
Official Documentation: docs.clawd.bot
GitHub Repository: github.com/clawdbot/clawdbot
Discord Community: Two Clawbot instances answer questions 24/7. Bugs get fixed while you’re reporting them.
SOUL.md: Edit this file to customize your personal AI assistant’s personality and behavior.
Your Action Plan: Getting Started Today
You now have everything you need to install a personal AI assistant.
Here’s your checklist:
- ✅ Install Node.js 22+
- ✅ Run
curl -fsSL https://clawd.bot/install.sh | bash - ✅ Run
clawdbot onboard --install-daemon - ✅ Get your Anthropic API key at console.anthropic.com
- ✅ Connect WhatsApp by scanning the QR code
- ✅ Send your first message
Total time: 30 minutes.
After that, you have a personal AI assistant that:
- Lives in your messaging apps (WhatsApp, Discord, Slack, etc.)
- Remembers everything you tell it
- Works while you sleep
- Costs less per month than a streaming service
- Runs on hardware you own and control
Welcome to having a true personal AI assistant. The 2026 version actually works.
Frequently Asked Questions
Q: Is Clawbot really open-source?
A: Yes, 99.99001% open-source. The founder kept one small file closed-source as a security measure and personal challenge.
Q: Can I use Clawbot on multiple devices?
A: Yes. Run separate instances or sync the same workspace across machines.
Q: What if I want to use a local AI model instead of paying for Claude?
A: Clawbot supports LM Studio for local models. Quality varies compared to cloud models, but it’s free.
Q: Will my data be private?
A: Yes. Everything runs on hardware you control. No data goes to Anthropic unless you explicitly send it through the API.
Q: Can I share my personal AI assistant with family members?
A: Yes. Configure separate agents with different permission levels.
Q: How often do I need to restart Clawdbot?
A: The daemon handles restarts automatically. Once running, it stays on unless you manually stop it.
Q: Why is everyone buying Mac Minis?
A: iMessage support, quiet operation, and low power consumption. But it’s not required—any hardware works.
Q: Is a personal AI assistant like this going to take my job?
A: Not directly. But it will compress workflows and make teams more efficient. The question isn’t whether AI changes work. It’s how you adapt.
What Comes Next
You’re now part of the wave of people building with personal AI assistants. The tools are evolving rapidly. The possibilities are expanding daily.
In six months, Clawbot will be significantly better than it is today.
In 12 months, this setup guide will feel quaint.
But the fundamental principle remains: a personal AI assistant that runs on your hardware, remembers your context, and executes real work is the direction the entire industry is moving.
You’re just getting there first.
Now go set up your personal AI assistant.
For more deep dives into the technology shaping our world, keep reading Welp Magazine, your source for the latest in tech, startups, and innovation.
