How do you structure Agentic OS for both Claude Code and Codex?

Hey everyone,

Is anyone here actively using an Agentic OS?

If so, I’m guessing most of you originally built it primarily around Claude Code. I’m currently planning to restructure mine a bit because it’s quite heavily Claude-focused, but lately I’ve been using Codex more and more since I find it faster and more token-efficient for certain tasks.

For those of you who use both Claude Code and Codex within the same Agentic OS, how have you structured your core setup and file tree so that both agents clearly understand what they’re supposed to do without getting in each other’s way?

Do you keep shared instructions somewhere model-agnostic and only use Claude/Codex-specific files for tool-specific behavior, or do you handle it differently?

Thanks!

10 comments

How do you structure Agentic OS for both Claude Code and Codex?