[JoseMar]

OSINT-with-LLM
local LLMs quietly changing how recon actually works

Hi guys, I was reading about a project called OSINT-with-LLM, and I almost skipped it. Looked meh. Another “AI” repo, right? But then it clicked and yeah, this is actually kind of important.

Not hype-important, real-important.

What actually happened here

So the idea is simple. You give it a target, email, domain, IP, whatever. It runs normal OSINT stuff, WHOIS, reputation checks, exposed services, threat intel sources. Nothing new. The difference is what happens after.

Instead of dumping raw output and calling it a day, it feeds everything into a local LLM and generates a report that explains what matters and why. Not “AI hacker”, not magic, just context. Honestly, that’s the part that surprised me.

And yeah, it runs locally. No cloud. No APIs. Your data stays yours. That alone already puts it in a different category for me.

Why people say it’s meh (and why that’s the point)

I’ve seen people say “this doesn’t do much”, and to be fair, they’re right. It doesn’t exploit anything. It doesn’t automate attacks. It doesn’t pretend to be an AI analyst god.

The LLM isn’t doing OSINT. It’s interpreting it. Like a junior analyst reading through noise before you do. That’s why it feels boring.

But that’s also why it matters.

This isn’t chatbot AI. This is AI as a component inside a workflow. Translating raw technical data into something a human can actually reason about.

That’s the takeaway I wanted to pass on. The interesting AI right now isn’t loud. It’s these quiet tools that sit in the background and just… help you think.

Link:

• https://github.com/mouna23/OSINT-with-LLM