Show HN: Agentpriv – Sudo for AI Agents

github.com

3 pointsnichkej3 comments

AI agents call tools autonomously, but some calls (delete_db, reset_state) shouldn't run unchecked.

agentpriv is a tiny permission layer: wrap any callable with allow/deny/ask and it gates execution before the function runs.

Zero dependencies, ~100 lines, works with any framework or plain Python. Happy to hear what's missing.

Open articleView on HN

Looks simple but very useful. This came across my screen at just the right time, have a few ideas that could do with something like this.

Glad it came at the right time. Honestly, even though the tool seems simple, I've been thinking of using it not only for raw access control, but to evaluate how different models behave with risky tasks - e.g. give it access to malicious tool but guarding it with 'deny'.

I'm curious to hear what your workflow ideas are.

I meant simple in a good way.

I'm utilising codex within an application for some simple inference stuff in read-only mode, but this gives me a way to easily wrap it in some guardrails.

Reply on news.ycombinator.com