With the rise of AI agents and recent advancements in agentic research, my co-founder and I wanted to explore whether we could develop a tool to automate parts of penetration testing. This led to the creation of TARS a few months ago.

TARS is our attempt to automate parts of cybersecurity penetration testing using AI agents. Currently, we leverage CrewAI as our agentic framework and GPT-4 as our primary model. We've integrated tools that allow TARS to utilize Nettacker, RustScan, ZAP, Nmap, and web browsing via the Brave API & Selenium.

Recently, we decided to make TARS accessible to everyone without any restrictions, so we open-sourced TARS under the MIT license. We see the value in TARS's ability to generate general reports from surface-level scans and its potential to perform unique attacks based on the model you're using.

Since TARS is in early development and still experimental, we'd greatly appreciate your feedback.

You can check out TARS here: https://github.com/osgil-defense/TARS

And view some demos of TARS here:

- Short Demo: https://www.youtube.com/watch?v=Sjw_gkSz6Lw

- Long Demo: https://www.youtube.com/watch?v=JSBVHl7PWek

Thank you for reading!