How do you couple them together efficiently? The nice thing about Codex or Claude is that the delegation or multi agent workflow capabilities are just built-in.

Do you link one with the other as a skill or mcp or so?

When I was going through this it was because OpenAI had defaulted to /fast mode with 2x token usage