Skip to content
Kitbay
Skill

Agent Cost Report

Apache-2.0Updated on Nov 18, 2026118,035GitHub starsIncludes scripts

A Claude-Mem skill that produces a cost and failure report on a user's coding-agent work, by default for the last seven full days.

What the skill teaches

A Claude-Mem skill that produces a cost and failure report on a user's coding-agent work, by default for the last seven full days. A bundled Python script reads token usage from local Claude Code and Codex transcripts, prices it at OpenRouter's public list prices fetched at run time (so dollar figures are labelled estimated), and uses Claude-Mem's memory search to label work items and cite evidence. The report asks what was finished, what each outcome cost, what was wasted on loops, rework or wrong turns, and whether unapproved actions were attempted. Output is an HTML report, an optional PDF from headless Chrome, and JSON and CSV files. Optional sources such as an OpenRouter key need the owner's approval each time.

Capabilities

  • Total token use from Claude Code and Codex transcripts
  • Estimate cost per completed piece of work
  • Classify wasted effort such as looping and rework
  • Write HTML, PDF, JSON and CSV reports