I built an autonomous AI pentesting agent for Kali Linux (because standard LLMs kept refusing to help)

Oveek Sur — Tue, 23 Jun 2026 06:57:11 +0000

Have you ever tried doing standard network reconnaissance or payload crafting using ChatGPT or Claude, only to be hit with: "I cannot assist with that"?

As a solo dev and security enthusiast, this ethical guardrail bottleneck was driving me crazy. I didn't want a malicious tool, I just wanted an AI that could actually help me automate my local Kali Linux workflows without constantly lecturing me.

So, I built FalconHunter.

What is it?
FalconHunter is an autonomous AI agent that connects directly to your Kali Linux box. It's not just a chat interface. It’s designed to run tools natively in your terminal, parse the live output, and decide on the next logical steps.

How it works under the hood:
Node.js Client: You run a lightweight Node.js agent on your machine that connects to the dashboard. No heavy installations.

Uncensored Models: It hooks into OpenRouter to access uncensored AI models, entirely bypassing artificial limitations.

Human-in-the-Loop Architecture: I didn't want the AI running rm -rf or aggressive exploits on its own. So, it explicitly pauses and asks for your approval before executing commands.

Why I'm sharing it here
I built this entirely bootstrapped to solve my own problems. I’ve just launched the MVP and I'm offering 15 free commands per day so the dev and security community can test it out.

I would absolutely love it if you guys could test the architecture, roast the code, and give me some raw feedback on how to handle large terminal outputs without blowing up the AI's context window.

🔗 Try it out here: https://falconhunter.lovable.app

Let me know your thoughts in the comments! 👇

DEV Community: Oveek Sur

I built an autonomous AI pentesting agent for Kali Linux (because standard LLMs kept refusing to help)