I’m trying to understand a specific multi-session Codex workflow.
When I have a few interactive Codex sessions open, I often switch away while one is working. When I return, the problem is not simply “did it finish?” It is:
-
which session produced this result?
-
what task and constraints did I give it?
-
what changed, and what should I review next?
Terminal tabs, worktrees, named sessions, and notes all help. But I still sometimes spend time reopening the wrong thread or rebuilding the context before I can review the result.
For full disclosure, I am building a small local macOS tool called Cairn around this narrow handoff: finished work can wait for review, then take me back to the originating session. It is not an agent manager, a remote-control app, or a request for Codex to add a feature.
I would like to learn from people who use Codex in real projects:
-
What does your last “come back and review” moment look like?
-
How many interactive Codex sessions can you realistically keep track of?
-
What do you use today to reconnect a finished result with its original context?
If you use Codex on macOS and this happens regularly, I am also looking for a few people to try the workflow on one real task. No call is required. I am mainly interested in what fails or feels unnecessary.