Agent Zero on Contabo: Ollama setup and security testing

This tutorial sets up Agent Zero on a Contabo VPS and uses it to test deliberately vulnerable web applications. The presenter installs the agent through its quick-start flow, creates web login credentials and opens its Docker-based interface through the server's IP address. The installation discussion also covers macOS, Linux and Windows options.

Model setup covers the main, utility and embedding roles. The presenter installs Ollama on the VPS, signs in and configures a cloud model, then adjusts the service configuration so Agent Zero's container can reach it. Although local models are discussed, this demonstration does not establish an offline inference setup. It shows a self-hosted agent connected to a cloud model through Ollama.

The security examples include a public lab and a custom Python web app on port 5000. When a port scan encounters an "Operation Not Permitted" error, the presenter shows Agent Zero writing Python code as a fallback. He reports SQL injection findings and checks a payload in the demo app. Further prompting leads to reported SSTI and IDOR findings, so the initial scan does not identify every planted flaw.

The walkthrough ends with a downloadable Markdown report containing proof of concept details and suggested fixes, a tour of the built-in browser and desktop, and HTTPS remote access through a Cloudflare tunnel.