Hey, thanks for the detailed and honest post, and for understanding that this isn’t about the date.
Basically, nothing has changed since June. Token usage metadata input and output counts still isn’t included in hook payloads, and I can’t share a committed roadmap or an ETA. I don’t want to give a vague “maybe”. As of today, this is a feature request we’re tracking, with no promised timeline.
On your two questions:
-
I can’t publicly share implementation details about why token fields aren’t available at hook time. This isn’t a brush-off, I just don’t want to speculate about internal mechanics in a public thread and mislead you. What I can say concretely is that the minimal version input and output counts per request in the existing payload, with no aggregation and no new API is exactly the shape I’ll pass to the team separately, since it’s much simpler than full token observability.
-
Honestly, I can’t confirm this in a committed roadmap in any form right now. If this is what you need to decide whether to commit to a workaround, it’s reasonable to assume there won’t be a built-in solution soon. If anything changes around token fields in hooks, I’ll post an update here.
Also, I want to call out the context shift you mentioned. The thread started as efficiency tuning for power users, but the recent posts including Gabriel are about attributing spend to projects and cost centers, which is more of a procurement use case. That’s a useful signal, and I’ll share it with the team in exactly those terms. It’s a different use case, and it changes the weight of the request.
A couple related threads on attribution, if you want to add your voice there too:
- We need a deterministic way to attribute Cursor token usage to local IDE sessions, features, and subagents
- How to obtain token usage per request?
I’ll also note your offer to test an early build. If that becomes an option, I’ll let you know in the thread.