Анализ расходов на AI-модели из логов

★ 7.5 · general

model-usage is a Claude Code skill that reads per-model cost data from CodexBar's local logs and summarizes spending broken down by Codex or Claude models. It runs through a bundled Python script (`model_usage.py`) with two modes: `current` — picks the most recent daily entry with the highest-cost model — and `all`, which produces a full breakdown across every recorded model. Output can be plain text or pretty-printed JSON; input can come from the CodexBar CLI, a file, or stdin. Note that CodexBar does not split token counts by model, so only cost figures are available per model. Useful for developers tracking which model drives the most spending or building scriptable cost analytics.