One of the unexpected joys of building a Personal AI Infrastructure is the moment you realize: this doesn’t have to live on just one machine.
My PAI runs on a Windows workstation at home. It has all my context, my skills, my workflows, my history. But I’m not always at that desk. Sometimes I’m on my phone. Sometimes my daughter wants to use her Chromebook for homework help. The question became: how do I access PAI from anywhere?
The Solution: Claude Code UI + VPN
The answer turned out to be surprisingly elegant. Two components make this work:
- Claude Code UI - A web-based interface for Claude Code
- VPN - Secure tunnel back to my home network (I use OpenVPN built into my TP-Link Deco router)
That’s it. No complex cloud deployments. No API proxies. No credential juggling. Just a web interface running on my home machine, accessible through a secure tunnel.
How It Works
Here’s the architecture:
┌─────────────────────────────────────────────────────────────────────┐
│ HOME NETWORK │
│ ┌─────────────────────────────────────────────────────────────┐ │
│ │ Windows Workstation │ │
│ │ ┌─────────────┐ ┌──────────────────────────────────┐ │ │
│ │ │ PAI │ │ Claude Code UI │ │ │
│ │ │ (.claude/) │◄───│ (Web Interface on port 3000) │ │ │
│ │ │ - Skills │ │ │ │ │
│ │ │ - Context │ └──────────────┬───────────────────┘ │ │
│ │ │ - History │ │ │ │
│ │ └─────────────┘ │ │ │
│ └────────────────────────────────────┼────────────────────────┘ │
│ │ │
│ ┌────────────────────────────────────┼────────────────────────┐ │
│ │ TP-Link Deco Router │ │ │
│ │ (OpenVPN Server Built-in) │ │ │
│ └────────────────────────────────────┼─────────────────────────┘ │
└───────────────────────────────────────┼─────────────────────────────┘
│
════════════════════╪════════════════════
SECURE VPN TUNNEL (Encrypted)
════════════════════╪════════════════════
│
┌───────────────────────────────┼───────────────────────────┐
│ │ │
┌────▼────┐ ┌─────▼─────┐ ┌──────▼──────┐
│ Phone │ │ Chromebook │ │ Laptop │
│ (Safari)│ │ (Browser) │ │ (Anywhere) │
└─────────┘ └────────────┘ └─────────────┘
The flow:
- Claude Code UI runs on my Windows workstation, serving a web interface
- My VPN (OpenVPN on my TP-Link Deco) creates a secure tunnel back to my home network
- From any device on the VPN, I can access
http://workstation:3000 - The web interface connects to my local Claude Code instance with all my PAI context
Why This Works So Well
Everything Stays Local
My PAI configuration, skills, history, and context never leave my home network. The VPN just provides a secure window into that environment. No cloud sync needed. No credentials stored on multiple devices.
Full PAI Experience
This isn’t a dumbed-down mobile version. When I connect from my phone, I get the exact same PAI that I use at my desk. Same skills. Same context. Same workflows. My AI knows who I am whether I’m at my desk or on the couch.
Family Access
My daughter can do homework on her Chromebook and get help from PAI. The Family skill knows about her interests and grade level. The Teaching skill (which I use for my university courses) adapts to help with elementary school questions too.
No Per-Device Setup
New device? Just install the VPN client and open a browser. That’s it. No Claude Code installation. No configuration sync. No API key management.
Setting It Up
1. Install Claude Code UI
Claude Code UI is an open-source web interface for Claude Code. On your PAI host machine:
# Clone the repository
git clone https://github.com/anthropics/claude-code-ui.git
# Install dependencies
cd claude-code-ui
npm install
# Start the server
npm run start
By default, it runs on port 3000. You can configure it to start automatically on boot.
2. Configure Your VPN
I use the OpenVPN server built into my TP-Link Deco mesh router. Most modern routers have VPN server functionality built-in:
- Enable OpenVPN server in your router’s admin panel (usually under Advanced > VPN Server)
- Export the .ovpn configuration file from the router
- Install OpenVPN Connect on your mobile devices
- Import the .ovpn file into the OpenVPN app
The Deco makes this particularly easy - it generates the config file with all the certificates included, so setup on each device takes about 30 seconds.
3. Access From Anywhere
From any device connected to your VPN:
- Open a browser
- Navigate to
http://[your-workstation-ip]:3000 - Start chatting with your PAI
That’s it. Your full Personal AI Infrastructure, accessible from your pocket.
Real-World Usage
Morning commute (phone): “What’s on my calendar today and what prep do I need?”
Daughter’s homework (Chromebook): “Can you help me understand fractions? Use examples with pizza slices.”
Coffee shop (laptop): “Let’s continue working on that email skill we started yesterday.”
Waiting room (phone): “Research the new Qualtrics XM Discover features for my Agilent project.”
In each case, PAI has full context. It knows my projects, my family, my teaching schedule, my work priorities. The device is just a window.
Security Considerations
A few things to keep in mind:
VPN is essential - Never expose Claude Code UI directly to the internet. Always access through VPN.
Device security matters - If someone gets access to a device on your VPN, they can access your PAI. Use strong device passwords.
Session awareness - Claude Code UI maintains sessions. Be mindful of leaving sessions open on shared devices.
Network segmentation - Consider putting your PAI host on a separate VLAN if you want extra isolation.
What’s Next
This setup has opened up possibilities I hadn’t considered:
- Voice interface - Adding a speech-to-text layer for hands-free PAI access
- Tablet dashboard - A dedicated tablet showing PAI status and quick actions
- Family profiles - Different skill configurations for different family members
The core insight is that PAI doesn’t have to be tied to one desk, one machine, one context. It’s your AI infrastructure, and you should be able to access it however works best for you.
This is part of my ongoing series on building and extending Personal AI Infrastructure. See also: PAI Explained: How Personal AI Infrastructure Works and Building Personal AI Infrastructure.