Dhruvit's profile on Product Hunt

About

I build infrastructure for Securing AI agents for Enterprises to use in Zest trust Arch. I care a lot about agent security: which MCP servers to trust, real isolation vs. containers, and seeing what your tools actually send. Building something in AI, devtools or security? Say hi, I'm always up to swap notes.

Links

kosa8

Badges

Gemologist

Tastemaker

Gone streaking 10

Gone streaking

View all badges

Maker History

kosa8Secure infrastructure for AI agents
Nov 2026

🎉

Joined Product HuntOctober 6th, 2026

Forums

p/kosa8-deleted-1337335

•

17d ago

Launching Tue Oct 20: try kosa8 early and tell me what breaks

We go live on Product Hunt on Tuesday. Before then, I'd love a few people to put the https://kosa8.com/ through its paces:

Post your rough edges, confusing docs or "why doesn't it do X" here. I'll fix what I can before launch and credit you in the changelog. I'll also be here all day on the 20th for questions.

p/kosa8-deleted-1337335

•

17d ago

Is Docker good enough isolation for AI-written code? Convince me either way.

Our take: containers share the host kernel, so for untrusted code a microVM is the safer wall, and kosa8's boot in about 0.5s, so speed isn't the trade-off it used to be.

The counterargument: gVisor and seccomp profiles go a long way, and VMs add complexity.

p/kosa8-deleted-1337335

•

17d ago

How do you decide which MCP servers to trust?

Every MCP server you add is code running with your permissions: your files, your network, your tokens. Most of us install them from a README and a star count.

Do you read the source? Pin versions? Run them in a container?