I got tired of paying for multiple SaaS subscriptions when everything could run in one place. Netlify, Supabase, Hostinger, n8n cloud — the monthly fees kept adding up, and managing four different dashboards felt like unnecessary overhead.
Before the AI era, setting up and maintaining a self-hosted VPS from scratch was something I deliberately avoided. The endless terminal configuration, SSL certificates, firewall rules, and deployment pipelines took too much time.
Last weekend, I decided to run an experiment: buy a fresh VPS and migrate every single project, site, and n8n workflow using AI only. Zero manual terminal work.
Setting Up the Server in Between Daily Tasks
Casually, between walking with my son, a quick trip to IKEA, and scrolling through Instagram, I bought a Hetzner VPS for around 6€ per month, installed Ubuntu, and gave Claude full terminal access.
First, we set up Coolify — an open-source, self-hosted PaaS alternative. I walked through the server structure with Claude so he could understand the architecture before touching any production files.
Then we started migrating:
- Drupal websites: Migrated from legacy shared hosting. There was one funny hiccup where Claude skipped a few images and accidentally pushed an old CSS file to production. When I pointed it out, he apologized and fixed the layout within two minutes.
- Modern JavaScript apps: Next.js and Astro projects deployed via Coolify without a single issue.
- n8n workflows: Exported all workflows, imported them into the self-hosted instance, and re-authenticated API keys.
The Real Numbers and Benefits
The outcome was immediate:
- Cost reduction: Instead of paying separate plans for hosting, databases, and workflow automation, everything runs on a single VPS for 6€ per month.
- Full architectural control: Any custom Docker container, MCP server, or Python script can be spun up in seconds.
- Direct AI operations: AI can inspect container logs, update configurations, and deploy new features directly when requested.
Safety and Backups
Giving an AI assistant terminal access carries real risks. A model could misconfigure permissions or overwrite critical volumes.
To make this setup safe, I established a strict multi-layer backup protocol:
- Daily full-server snapshots
- Automated per-project volume backups
- Version-controlled git repositories for all source code
Even if an agent makes a mistake, restoring the entire server takes less than five minutes.
If you are looking to simplify your infrastructure or automate repetitive client workflows, we can help you design a lean, reliable setup. Check out our automation services or get in touch for a free workflow audit.
