I use Codex on multiple computers and also on a server, and I think there is a big opportunity for Codex sessions on different machines to communicate with each other.
Example workflow:
- I have Codex running on my development PC.
- I also have Codex running on my server.
- The server has an API problem and starts returning HTTP 500 errors.
- From my development PC, I want to be able to tell Codex:
“Ask the Server Codex why the API is returning 500 errors.”
The Server Codex could then inspect its logs, stack traces, service status and other relevant information, and send the findings directly back to the Dev Codex.
The Dev Codex could fix the code and then tell the Server Codex:
“The fix is ready. Please pull/test commit abc123.”
The Server Codex could test the fix and return the result.
This would make multiple Codex sessions behave much more like a real team of agents instead of isolated chats.
It would also be useful for one Codex session to hand work over to another:
“I finished the database migration. Tell the testing Codex it can continue.”
Each device/session could have explicit permissions such as:
- read logs
- inspect files
- run tests
- execute commands
- read-only access
- full access
- send messages to another authorised Codex session
Ideally this could work between Codex sessions connected to the same OpenAI account or workspace.
I think this could be especially useful for developers running separate development, testing, database and server environments.