Commit

36220ac

Some thoughts raised in my mind this morning, let me tell you. We embed AI models in our applications and give them tools to call. Being careful with these tool is what makes the AI kind of sandboxed. But now, models are smart enough to search for vulns, they may try to run…

Younes @younes-io4h762afe1

run the agents in a sandbox with kernel-level control : https://github.com/nolabs-ai/nono

GitHubGitHub - nolabs-ai/nono: agent runtime security - zero trust, zero setup, zero latency agent sandboxingagent runtime security - zero trust, zero setup, zero latency agent sandboxing - nolabs-ai/nono

Julien Bianchi @jubianchi3h4205c76

Thanks, did not know about this project, will look. Reading the readme I don’t really see how I would implement this with an AI embeded in my app (whatever the app: web, native, …). Looks great for local agent though

Younes @younes-io3h1/23c3cd91

i think wrapping the app with nono run and pointing the SDK at nono's proxy should work today.. maybe you can try something like and see: nono run --profile my-app.json -- node server.js

Younes @younes-io3h2/2

if you want only tool execution, they have rust and python bindings

08:00 · Oct 27, 2026commit 36220ac

Julien Bianchi @jubianchi3h

will have a look at all that, thanks for the hints btw