TL;DR — Get Thunderbolt Running in 5 Minutes

1. Get a Netcup VPS 2000 G12 (8 vCPU, 16 GB RAM, 512 GB NVMe — €14.33/month)

Use one of these coupon codes for 1 month free:

5800nc17718015231
5800nc17718015230
5800nc17755880760

2. Deploy with Docker Compose on your server:

apt update && apt install -y docker.io docker-compose git
cd /opt
git clone https://github.com/thunderbird/thunderbolt.git
cd thunderbolt/deploy
cp .env.example .env
docker compose up --build

3. Access Thunderbolt at http://your-server-ip:3000 and login with Keycloak credentials ([email protected] / demo)

Need more power? Check the full server comparison below — including Root Servers with dedicated CPU cores and up to 2 months free.

Introduction: Own Your AI Infrastructure

The AI revolution has brought powerful tools like ChatGPT, Claude, and Gemini to the forefront. But there's a critical problem: your data flows through third-party servers, and you're locked into each provider's ecosystem. For enterprises and privacy-conscious users, this creates vendor lock-in, compliance challenges, and data sovereignty concerns.

Mozilla's MZLA Technologies has a solution. Thunderbolt is an open-source "sovereign AI client" that puts you in control of your AI infrastructure. Developed with funding from Mozilla, Thunderbolt lets you choose your own models, host your own data, and eliminate vendor lock-in — all while maintaining enterprise-grade security.

This guide covers everything you need to deploy Thunderbolt on a Netcup VPS, from understanding what Thunderbolt is to production-ready deployment.

What is Thunderbolt?

Thunderbolt is an open-source, cross-platform AI client developed by Mozilla's MZLA Technologies (the team behind Thunderbird email). Positioned as a "sovereign AI client," it gives users complete control over their AI infrastructure rather than relying on cloud-based services.

Core Features

Multi-Platform Access Thunderbolt runs on web, iOS, Android, Mac, Linux, and Windows. A single React codebase powers all platforms through Tauri, ensuring consistent experience across devices.

Model Agnostic Connect to any OpenAI-compatible model provider. Use frontier models from OpenAI, Anthropic, Mistral, or local models via Ollama or llama.cpp. The inference proxy handles routing, rate limiting, and API key management.

Enterprise-Grade Security Optional end-to-end encryption means data is encrypted before leaving your device. The server stores only ciphertext. Device-level access controls add another security layer.

Self-Hosted Deployment Run the entire stack locally — backend API, PostgreSQL database, PowerSync for real-time sync, and Keycloak for authentication. Everything stays on your infrastructure.

Haystack Integration Built on Deepset's Haystack framework, Thunderbolt integrates with enterprise data through industry-standard protocols like MCP (Model Context Protocol) and ACP (Agent Client Protocol).

Why "Sovereign" AI Matters

Sovereign AI means:

• Data stays local: No third-party servers see your prompts, responses, or files
• Model choice: Switch providers without migration pain
• Compliance: Meet GDPR, HIPAA, or sector-specific data requirements
• Cost control: Flat hosting fees vs. per-token pricing

Current Development Status

Thunderbolt is under active development. As of 2026, it's undergoing a security audit and preparing for enterprise production readiness. The code is available for self-hosting and testing, but production deployments should monitor official updates.

Why Host Thunderbolt on a VPS Instead of Locally

Running Thunderbolt locally works for personal testing, but a VPS unlocks its full potential.

Always-On Availability

Your home internet goes down; your server sleeps. A VPS in a professional data center with redundant power and networking keeps Thunderbolt running 24/7, accessible from any device anywhere.

Enterprise Data Protection

AI applications often handle sensitive data. A VPS in a German data center provides GDPR compliance, data sovereignty, and professional security controls — essential for enterprise deployments.

Dedicated Resources

AI workloads need consistent CPU, RAM, and I/O. A VPS gives Thunderbolt dedicated resources that won't compete with your other applications.

Team Collaboration

A centrally hosted Thunderbolt instance serves multiple team members simultaneously — impossible with a local install. Everyone gets the same AI assistant with shared context.

Static IP and Domain

Production AI tools need consistent endpoints. A VPS provides a static IP and the ability to set up proper domain names with SSL certificates.

Why Netcup: The Ideal Host for Thunderbolt

Netcup offers several advantages specifically relevant for Thunderbolt hosting.

German Data Privacy

All Netcup data centers are in Germany and Austria, subject to strict GDPR. Your AI data stays in the EU — critical for compliance-sensitive deployments.

Price-to-Performance

Netcup delivers 60-80% cost savings compared to AWS, Google Cloud, or Azure. For always-on AI infrastructure, this adds up significantly.

Enterprise Hardware

AMD EPYC processors, DDR5 ECC RAM, and NVMe SSDs provide the performance AI workloads need without premium pricing.

Included Features

DDoS protection, backups, snapshots, and API access are included. No surprise charges for bandwidth or basic features.

Transparent Pricing

The price you see month one is the price you pay ongoing. No introductory rate spikes after the first billing cycle.

Easy Scaling

Upgrade CPU, RAM, and storage through the control panel. Scale from a small VPS to a dedicated Root Server as needs grow.

Choosing the Right Netcup Server for Thunderbolt

Thunderbolt runs on Docker Compose with multiple containers. Here's how Netcup's offerings match Thunderbolt requirements.

Container Resource Requirements

The standard Thunderbolt stack includes:

• Frontend: nginx serving React SPA
• Backend: Bun API server with auto-migrations
• PostgreSQL: Database with WAL replication
• Keycloak: OIDC authentication
• PowerSync: Real-time sync service
• MongoDB: PowerSync storage backend

Minimum recommended: 4 vCPU, 8 GB RAM Recommended for production: 8 vCPU, 16 GB RAM

VPS (Virtual Private Server) — Recommended for Most Users

Netcup's VPS line uses shared hardware with guaranteed minimums. Ideal for personal use, development, and production workloads.

VPS 1000 G12 — Development & Testing

Best For:

• Development and testing Thunderbolt
• Single-user personal use
• Learning Docker Compose
• Light AI workloads

Coupon Codes (1 Month Free):

5799nc17755735170
5799nc17755734540
5799nc17755732650

VPS 2000 G12 — Recommended for Production

Best For:

• Production Thunderbolt deployment
• Teams of 2-5 users
• Running multiple concurrent AI sessions
• Production with PostgreSQL and PowerSync

Coupon Codes (1 Month Free):

5800nc17718015234
5800nc17718015233
5800nc17718015232

VPS 4000 G12 — High Performance

Best For:

• High-throughput AI processing
• Multiple concurrent users
• Heavy data processing workloads
• Running Thunderbolt alongside other services

Coupon Codes (1 Month Free):

5801nc17718015214
5801nc17718015213
5801nc17718015212

Root Servers (RS) — Dedicated Performance

Root Servers provide dedicated CPU cores with no resource sharing. Choose for production workloads requiring consistent, predictable performance.

RS 1000 G12 — Entry Dedicated

Best For:

• Production Thunderbolt with guaranteed performance
• Latency-sensitive AI workloads
• Consistent CPU availability

Coupon Codes (2 Months Free):

5159nc17718015444
5159nc17718015443
5159nc17718015441

RS 2000 G12 — Production Dedicated

Best For:

• Teams running Thunderbolt in production
• Sustained AI workloads
• Running multiple services alongside Thunderbolt

Coupon Codes (2 Months Free):

5160nc17718015411
5160nc17718015414
5160nc17718015413

RS 4000 G12 — Enterprise

Best For:

• Enterprise-grade Thunderbolt deployments
• Large teams with concurrent users
• Maximum performance without resource contention

Coupon Codes (2 Months Free):

5161nc17718015391
5161nc17718015390
5161nc17718015394

RS 8000 G12 — Maximum Power

Best For:

• Large-scale AI operations
• Multiple simultaneous AI agents
• Heavy data processing

Coupon Codes (2 Months Free):

5162nc17718015362
5162nc17718015361
5162nc17718015360

VPS vs Root Server: Which Should You Choose?

Our recommendation: Start with a VPS 2000 G12 using coupon 5800nc17718015231 to test your setup. Once ready for production, upgrade to a VPS 4000 G12 or move to an RS 2000 G12 for dedicated performance.

Quick Start Guide

Prerequisites

• A Netcup VPS or Root Server (VPS 2000 G12 recommended)
• SSH access to your server
• Basic command line knowledge

Step 1: Provision Your Netcup Server

1. Sign up at netcup.com
2. Order a VPS 2000 G12 (or higher)
3. Use coupon code 5800nc17718015230 for 1 month free
4. Choose Ubuntu 24.04 LTS as your OS
5. Select a data center location (Vienna, Nuremberg, or Amsterdam recommended for EU users)

Step 2: Connect to Your Server

ssh root@your-server-ip

Step 3: Install Docker and Dependencies

apt update && apt install -y docker.io docker-compose git ca-certificates curl

Step 4: Clone Thunderbolt Repository

cd /opt
git clone https://github.com/thunderbird/thunderbolt.git
cd thunderbolt/deploy

Step 5: Configure Environment

cp .env.example .env
nano .env

Update key variables:

• Set your domain if needed
• Configure authentication settings

Step 6: Start Thunderbolt Stack

docker compose up --build -d

This builds and starts all containers:

• Frontend (nginx)
• Backend (Bun API)
• PostgreSQL
• Keycloak
• PowerSync
• MongoDB

Step 7: Verify Deployment

Check container status:

docker compose ps

View logs:

docker compose logs -f

Step 8: Access Thunderbolt

Default Keycloak user: [email protected] / demo

Step 9: Configure Model Provider

In the Thunderbolt UI:

1. Go to Settings
2. Add your model provider API key
3. Choose your preferred model (Ollama for local, OpenAI-compatible provider for cloud models)

Production Configuration

SSL/TLS with Let's Encrypt

For production, enable HTTPS:

apt install certbot python3-certbot-nginx
certbot --nginx -d thunderbolt.yourdomain.com --redirect

Set up automatic renewal:

certbot renew --dry-run

Firewall Configuration

ufw allow 22/tcp    # SSH
ufw allow 80/tcp    # HTTP
ufw allow 443/tcp   # HTTPS
ufw allow 3000/tcp  # Thunderbolt
ufw enable

Data Persistence

Thunderbolt data is stored in Docker volumes. For production, configure regular backups:

docker volume ls | grep thunderbolt

Resource Limits

Prevent container resource exhaustion:

nano /etc/docker/daemon.json
{
  "default-ulimits": {
    "nofile": {
      "Name": "nofile",
      "Hard": 65536,
      "Soft": 65536
    }
  }
}
systemctl restart docker

Security Best Practices

SSH Hardening

Disable password authentication and use SSH keys:

nano /etc/ssh/sshd_config
PasswordAuthentication no
PubkeyAuthentication yes
systemctl restart sshd

Keycloak Security

Change the default admin password immediately after first login. Configure proper realm settings for production.

SSL/TLS

Always use HTTPS in production. Let's Encrypt provides free certificates.

Regular Updates

cd /opt/thunderbolt
git pull
docker compose build
docker compose up -d

Backups

Automate regular backups of the PostgreSQL and MongoDB data volumes.

Monitoring

Set up logging and monitoring:

docker compose logs -f --tail=100

Cost Comparison: Netcup vs Major Cloud Providers

Monthly Cost (approximate, excl. VAT)

Annual Savings

• Netcup VPS 2000: €172/year
• AWS equivalent: €1,080/year
• Annual savings: €908+

Use Cases

Enterprise AI Assistants

A company deploys Thunderbolt internally for employees to interact with AI models. All data stays on company infrastructure, meeting compliance requirements.

Privacy-Focused AI

Healthcare or legal organizations use Thunderbolt to process sensitive data with AI. Data never leaves the organization's servers.

Developer AI Workflows

Development teams integrate Thunderbolt into their workflows. Code reviews, documentation, and testing assistance with full audit trails.

Research Institutions

Universities run Thunderbolt for research AI workloads. Flat hosting fees vs. per-token API costs maximize research budgets.

Conclusion

Thunderbolt represents the future of sovereign AI — giving users control over their AI infrastructure rather than relying on cloud providers. Running Thunderbolt on Netcup combines the best of both worlds: enterprise-grade infrastructure at a fraction of cloud costs, with German data privacy protection.

To get started:

1. Pick a server plan from the recommendations above
2. Use a coupon code from this guide for free month(s)
3. Deploy with Docker Compose
4. Start building your sovereign AI infrastructure

Whether you're an enterprise needing compliance, a developer wanting full control, or a privacy-conscious user, Thunderbolt on Netcup delivers.

Start with a VPS 2000 G12 using coupon 5800nc17755880760 and scale from there.

This guide is intended to help you understand and deploy Thunderbolt on Netcup. Thunderbolt is under active development — always refer to the official Thunderbolt and Netcup documentation for the most current information.