I recently found the time to sit down with Claude Code [1] from Anthropic to get a better understanding of why friends and colleagues were so enthusiastic about AI-assisted coding. Claude Code is an agentic coding tool that, according to the vendor, "runs in a terminal/IDE, understands the full codebase and can act autonomously toward a goal." In other words, we have already moved far from just chatting to AI; instead we have given it agency to act on our behalf.

Working in cybersecurity, much of my focus has been on defending against attacks for clients, but over the last few years I have significantly improved my offensive security knowledge. And, sitting down with Claude Code, the first thing I thought of was chaining multiple security tools together to see what was possible.

In this article, I will demonstrate how to get started with Claude Code and apply it to an example of ethical hacking. The following approach will build a platform for Claude so it can send its agents off to pentest any target that I point it at, simply by typing the run command. In other words, I will set up pentest infrastructure using Claude to help with the Python coding, plus the integration of the tools involved. My involvement, as part of the assisted coding work, will be to make sure Claude Code is configured in a way that means I get the most reliable and repeatable results. I think this is a clever route to take because I get the benefits of running multiple agents via Claude to perform simultaneous pentests, and the configuration is fluid enough that I can easily make changes via Claude's CLI using chatbot-like instructions.

[...]