Skip to content
llm-usage-metrics

Overview

v0.6.0

Know what your
AI coding costs.

Local-first usage metrics for every AI coding agent. Token counts, real pricing, Git‑correlated ROI — sub‑second reports.

$ npm i -g llm-usage-metrics
Get started GitHub
llm-usage daily 
$ llm-usage daily

  Period      Source   Models              Input    Output   Total      Cost
  ──────────  ───────  ──────────────────  ───────  ───────  ───────  ──────
  2026-03-02  pi       • claude-sonnet-4   142,319   38,104  180,423  $1.57
  2026-03-02  codex    • claude-sonnet-4    98,712   21,401  120,113  $1.02
  2026-03-02  gemini   • gemini-2.5-pro     67,241   15,832   83,073  $0.44
  2026-03-02  codex    • o3                 31,049    8,198   39,247  $0.41
  ──────────  ───────  ──────────────────  ───────  ───────  ───────  ──────
  ALL         TOTAL    4 models            339,321   83,535  422,856  $3.44

Blazing fast.

Benchmarked against ccusage on real production data.

4.6×
faster cold start
3.6s vs 16.8s
22×
faster with cache
0.7s vs 17.0s
<1s
cached reports
sub-second
View full benchmarks →

Four commands. Full visibility.

Usage rollups, Git-correlated efficiency, pricing optimization, and daily trend visibility.

📊

Usage reports

Aggregate token counts and costs across all agents. Daily, weekly, or monthly. 

llm-usage monthly \
  --provider openai
⚡

Efficiency

Correlate LLM spend with Git outcomes. $/commit, $/1k lines, tokens per commit. 

llm-usage efficiency monthly \
  --repo-dir .
📈

Trends

Track daily cost or token movement over time. Combined view or source-by-source. 

llm-usage trends \
  --metric tokens
🔬

Optimize

Replay your token mix against candidate models. Find cheaper alternatives instantly. 

llm-usage optimize monthly \
  --candidate-model gpt-4.1

Every agent. Zero config.

Auto-discovers session data from five AI coding tools.

pi codex gemini droid opencode
100% local-first
0 telemetry
3 output formats
O(1) parse cache