News for hackers. Stuff that matters.

vinaypanghal

Show HN: Driftcop – Open-source CLI SAST for "MCP rug pull attacks in AI Agents"

https://github.com/sudoviz/driftcop

Hi HN! We just open-sourced Driftcop, a security tool for people building AI agents with external tools via MCP. Driftcop continuously checks that the tools your AI agent relies on haven’t changed or drifted in unsafe ways. The motivation came from recent findings that AI agents can be quietly compromised via their tools – e.g. a tool that was useful and benign yesterday could auto-update into...

( Read More... | 1 comment | 2 points | on HN )
Re:Show HN: TuringMind AI deeply understands your codebase on Thursday November 07, 2024 @04:47PM (#42081449)
So, I’ve been thinking a lot about how we interact with code lately. If you’re like me, you’ve probably spent hours scrolling through repositories, trying to figure out why a piece of code exists or what the original dev was thinking. And yeah, I’ve dumped plenty of snippets into ChatGPT, hoping for a shortcut. It works... kind of. But it’s also messy, disconnected, and doesn’t really "get" the big picture.

That’s why I started working on TuringMind.ai It’s not just another tool to throw your code into—it’s something I always wished I had: an AI that can reason about entire repositories. Imagine connecting your repo, asking a question like, "Why was this function implemented this way?" or "What’s the relationship between these files?" and getting answers that actually make sense, without copy-pasting a single line.

I built TuringMind.ai because I felt there was a gap in how we, as developers, engage with our own (or others’) code. It’s about more than just quick fixes—it’s about understanding. Whether you’re onboarding to a new repo, troubleshooting an issue, or just exploring, there’s so much more we could do if the tools really understood the context of our work.

We’re starting small, but the goal is big: to make it feel like your codebase has a built-in teammate who’s always there to help.

I’d love to know—what’s your biggest frustration when working with a new repo? Or that one feature you wish existed? Let’s talk about it. Maybe it’s something TuringMind can solve.

I'm trying to get insights into the kind of problems AI copilots are still unable to solve.
We've built a tool to do the same. We started with a use case of triaging CVEs and SAST findings but our focus has always on accuracy of results. Its been pretty challenging to make a highly accurate reasoning and retrieval engine and I don't think the problem is already solved yet.

In your opinion, is there anything that any of these tools don't already do well?

That’s a great list—thanks for sharing! Tools like Github CoPilot have been game-changers for me when it comes to simple tasks codebases. I’ve heard a lot about GitLoop recently, curious to hear your experience with it—does it handle large repositories well?

Show HN: Building an AI That Understands Your Code

A couple of years ago, I noticed a gap in developer workflows. We have amazing tools for writing and shipping code, but understanding someone else’s code—especially in large repositories—often feels like solving a mystery without a map. What if there was an AI that could explain codebases as easily as a teammate would?That idea led to TuringMind, an AI-powered code chat platform that helps...

( Read More... | no comments | on HN )