FAQ
Is the dollar figure my actual bill?
Section titled “Is the dollar figure my actual bill?”No. Claude Code transcripts contain token counts but not billed amounts, so every dollar figure is estimated from tokens × list price. See the accuracy contract.
Why is my token count so large?
Section titled “Why is my token count so large?”Agentic coding re-reads the conversation on every turn, and most of that re-reading comes from the prompt cache. Cache reads usually make up more than 90% of billable tokens. The Context hygiene page shows how much of your spend happened at large context sizes.
Does it need an API key or an account?
Section titled “Does it need an API key or an account?”No. It reads files that are already on your disk. An API key is optional and only used if you want to pull billing figures from a provider.
Does it send my prompts anywhere?
Section titled “Does it send my prompts anywhere?”No. The server only listens on 127.0.0.1. The only outbound request is a daily npm version check, which you can turn off. See Privacy & data.
My numbers dropped a lot after upgrading to 0.8.0. Why?
Section titled “My numbers dropped a lot after upgrading to 0.8.0. Why?”Version 0.8.0 fixed a bug where some requests were counted two or three times, corrected two wrong list prices, and stopped applying the long-context surcharge where it didn’t apply. On the author’s own data, estimated spend went from about $17,000 to about $3,400. The warehouse rebuilds itself automatically after the upgrade.
I upgraded to 0.9 or later. Do I need to do anything?
Section titled “I upgraded to 0.9 or later. Do I need to do anything?”Click ⟳ Sync once, or start with --rebuild. That picks up Claude desktop (Cowork) sessions and the newly priced Claude 4.x and Haiku 3.5 models.
Is it affiliated with Anthropic?
Section titled “Is it affiliated with Anthropic?”No. “Claude” and “Claude Code” are trademarks of Anthropic, PBC, used here only to describe what the tool reads.