I'd like to build a security agent of some kind, an agent that talks to an internal...

I'd like to build a security agent of some kind, an agent that talks to an internal private hosted llm and actually takes actions, runs code