How can we clearly see token usage per Codex conversation/task, especially in Codex web?
OpenAI Developer Community
May 22, 2026
If you need to hardwire things to that level, I’d recommend using the app-server in the Codex SDK.
It will give much more information you don’t see in the IDE/App and some that don’t even go to the .jsonl, plus you can control the model, get usage per request, etc.
Discussion in the ATmosphere