Observed arrival · 2026-09-16
quota-autopsy — where did your Claude Code quota go?
A local command-line tool that analyzes Claude Code session transcripts to identify expensive calls, cache failures, repeated reads, and other sources of token waste.
Field notes
The tool parses the session transcripts Claude Code already writes, then separates fresh input, generated output, cache writes, and cache reads before looking for costly patterns. Its sample report covers 46 sessions and 1,162 API calls, including repeated file reads, oversized tool output, and approximately $10.47 in estimated avoidable spend. The page also says duplicate rows are collapsed by message ID and that the analysis has zero runtime dependencies.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue