Tech • AI • Robotics • Game
Claude + VPS: Put your apps online without touching anything technical
6/10TL;DR
A new workflow using Claude Code, the Hostinger plugin and an MCP connector allows non-specialists to deploy websites and apps on a VPS with minimal manual server administration.
KEY POINTS
VPS hosting positioned as the main bottleneck
Building small websites, internal tools, association pages or niche mobile web apps has become easier with generative AI, but putting those projects online remains a major hurdle. The recurring obstacles are classic infrastructure tasks such as DNS, SSH, HTTPS, domain registration and server configuration, which are often outside the skill set of non-admin users and even many developers.
Hostinger VPS offered as the low-cost base layer
The setup described relies on a Hostinger VPS, presented as available for roughly 7 euros per month after promotional discounts, with a 30-day money-back guarantee and a free domain name for one year. A mid-range KVM 2 plan is highlighted as sufficient for many personal apps, websites and lightweight SaaS projects, with the option to upgrade later if traffic grows.
Ubuntu is the only manual server choice
The recommended manual configuration is intentionally minimal: choose a server location, install the latest version of Ubuntu, and stop there. No direct use of the control panel is required afterward, with server management delegated to AI rather than handled through conventional admin dashboards.
Claude gains direct control through plugin and connector
The core of the method is the official Hostinger plugin for Claude plus a matching Hostinger MCP connector. Once both are activated in the desktop app, Claude can inspect domains, identify VPS assets, manage settings and interact with the hosting environment directly, removing the need for most command-line work.
AI handled domain registration and DNS changes
After connection, Claude was able to detect that a free domain was still available and register renaud-decode.tech automatically after an unsuccessful attempt to claim a .org address. It then configured a redirect so that the new domain pointed to renaud-decode.fr, and proposed the use of subdomains such as app, tool or training for additional projects.
One VPS can host multiple projects
The approach emphasizes that a single VPS can carry multiple services, from websites to automation tools such as n8n, cloud services and custom web applications. By using one domain plus multiple subdomains, users can publish several independent projects without buying separate infrastructure for each one.
A translation app was built and deployed as a test case
To demonstrate the workflow, a mobile-friendly real-time translation web app was created for a trip to San Francisco and Microsoft Ignite. The app was designed to capture speech, translate conversations in real time, store the user’s OpenAI API key locally in the browser rather than in a database, and remain usable on a smartphone.
Prompt structure was treated like product specification
The application was generated from a five-part prompt covering the final goal, expected features, desired user experience, technical constraints and deliverables. That structure was used to keep the build focused on practical use, including one-handed smartphone operation, translation quality and safe handling of credentials.
Deployment was executed without direct sysadmin work
In a fresh Claude session, the completed local project directory was handed over with a request to deploy it on the Hostinger VPS under trad.renaud-decode.tech. Claude proposed a deployment plan, created temporary SSH access, reviewed resources and security settings, configured DNS, virtual hosts, HTTPS, and a rollback point, then completed the publication after approval.
The published app is already reachable online
The final result was a live web application at trad.renaud-decode.tech, accessible from a smartphone and ready for real-time interpretation use. The demonstration framed this not as a unique project but as proof that the technical barrier to shipping personal or commercial tools can be reduced to a conversation with an AI assistant.
CONCLUSION
The process points to a shift in web deployment: AI is no longer limited to writing code, but is increasingly taking over the operational layer that once blocked many projects. If that model proves reliable at scale, launching small apps and websites could become far more accessible to non-technical creators.
Ask a question