First Time Deploying to Production with AI Agents: Testing Cursor on Azure (pdub.click)

A developer tested Cursor — an AI agent — by deploying a YOURLS URL shortener to a fresh Azure VM and reports it handled the entire end-to-end production workflow autonomously. Cursor connected over SSH, ran installation commands, installed dependencies, configured Apache virtual hosts, set up a MySQL database, managed file permissions and security settings, and provisioned SSL certificates. When asked to add a custom plugin that prefixes short URLs with dates, the agent wrote a working plugin on the first try. The full deployment and customization took about 15 minutes (versus roughly an hour manually), and the live short link and a scrubbed transcript of the session are available alongside a step-by-step tutorial. This showcases a practical milestone for AI agents in DevOps: they can perform complex, privileged server operations and make sensible configuration decisions, significantly speeding up routine infra and deployment tasks. Key technical implications include improved productivity and lower barrier-to-entry for small deployments, but also renewed emphasis on auditability, security and trust (credential handling, change verification, rollback safety). For teams considering agents in production, the takeaway is that agents can be effective helpers for repeatable ops — provided you enforce strict logging, least-privilege credentials, and human review of critical changes.