Launching today

trove
token savings you can verify
5 followers
token savings you can verify
5 followers
Find out where your Claude Code tokens go. Trove quietly tests each fix on a copy of your sessions, only switches it on if it really saves, and undoes it if it stops working. Open source.






Free
Launch Team

AWS Activate CreditsCredits that fuel your startup’s next milestone
Promoted
Hi Product Hunt! Jaclyn here 👋
If you use Claude Code a lot, you already know where the tokens go: tool output, noisy logs, the same file read again. And there's no shortage of fixes. Collapse the repeats, clean up the noise, cache more, or send the easy turns to a cheaper model.
The hard part is knowing if a fix actually helped. Did it save anything after its own overhead? Did it break the prompt cache? Did your agent miss the one error line it needed, or do worse on the cheaper model? Most tools just turn the optimization on and show you a number.
That's where Trove comes in. Every fix runs on a copy of your real sessions first and records what it would have saved. It only gets switched on if the evidence is there, only between sessions, and it switches itself back off if the saving goes negative. Every number is labeled as measured or estimated, so you know what you're looking at.
We held ourselves to the same standard: our benchmarks are pre-registered and published in full, including what didn't work. But you don't have to take our word for it. Trove measures what each fix would save on your own sessions before it turns anything on.
Trove is free, open source and runs on your machine. Nothing is sent anywhere by default, and you can undo everything it changed with a single command.