I have recently been working on an operator to run and manage sandboxed operations on the OpenAI-managed API.
It's coming out of building and running in practice from my last consultancy assignment.
As a Platform engineer, I was fascinated by how I could leverage my development team, where they (developers) focus on building skills in their repo oriented around tasks like code review, RCA investigation or anything where they confidently point their terminal agent, and I provide the declarative infrastructure to let those agents run whenever they want (event-based, ad hoc, hourly, etc.) with governance, cost control and guardrails.
Manage Agents API is awesome because it separates the descion/interligence and tool invocation environment (self-managed).
I thought the community would provide me with feedback and guidance to further pursue this path.
https://github.com/softcane/codex-session-operator
submitted by