The author developed a tool called `agent-cost` to address the limitations of existing usage trackers for AI models like Claude Code and Codex. Unlike broader CLIs or dashboards, `agent-cost` focuses on providing auditable, machine-readable facts about AI usage directly from local logs, without network calls during runtime. This approach separates observable session usage from the separate claim of task attribution, preventing the inaccurate apportionment of costs based solely on elapsed time or commit counts. AI
IMPACT Provides a more granular and auditable method for tracking AI model costs, separating usage from task attribution.
RANK_REASON The item describes a new software tool for tracking AI model usage.
Read on dev.to — Claude Code tag →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
