npx code-usage

See how much your AI coding actually costs — across all your tools, in one place.

code-usage reads local session files, calculates API-equivalent costs, and builds a self-contained HTML dashboard. No accounts, no APIs, no data leaves your machine.

Why code-usage?

Most usage trackers cover a single tool. If you use Claude Code and Codex and Amp, you need separate tools for each.

code-usage gives you one unified dashboard with combined views, stacked charts, and side-by-side comparisons — for every AI coding agent you use.

Supported Tools

| Tool | Status | |------|--------| | Claude Code | Supported | | Codex CLI | Supported | | OpenCode | Supported | | Amp | Supported | | Pi-Agent | Supported |

Only tools with local data appear in the dashboard — no empty tabs, no clutter.

Features

• Combined + per-agent views — stacked charts showing cost breakdown across all your tools
• Per-model costs — see exactly which models eat your budget (Opus 4.6 vs Sonnet 4.5 vs GPT-5.3)
• Per-project breakdown — donut chart with cost/session data per project
• Daily & weekly trends — spot patterns in your usage over time
• Date range filtering — This Week, This Month, Last 30/90 Days, All Time
• Dark mode — System / Light / Dark theme toggle, persisted across sessions
• Token breakdown — input, output, cache read/write, reasoning tokens with tooltips
• Usage streaks — see your consecutive days of AI coding activity
• JSON export — --json flag for scripting and automation
• 100% offline — everything runs locally, nothing is uploaded

Quick Start

npx code-usage

Or install globally:

npm install -g code-usage

That's it. If you have session data from any supported tool, the dashboard opens automatically.

CLI Options

code-usage                  # Open dashboard in browser
code-usage --no-open        # Generate HTML without opening
code-usage --json           # Print structured JSON to stdout
code-usage --range 30d      # Filter: 7d, 30d, 90d, all
code-usage -v               # Version
code-usage -h               # Help

Output

~/.code-usage/current/code-usage-dashboard.html   # Interactive dashboard
~/.code-usage/current/openusage-data.json          # Raw data snapshot

How It Works

1. Reads local session files from each tool's standard data directory
2. Calculates token counts and API-equivalent cost estimates per model
3. Aggregates by day, week, project, and model
4. Builds a single self-contained HTML file with all data inlined
5. Opens it in your browser

Pricing is based on published API rates. Fallback pricing powered by LiteLLM. If you're on a subscription plan (Claude Max, Codex Pro), actual billed cost may differ.

Privacy

All data stays on your machine. code-usage only reads standard local session files — no .env, no API keys, no network calls.

Development

npm install
npm run lint              # Biome
npm test                  # Vitest (53 tests)
npm run build:dashboard   # Rebuild template
node bin/code-usage.js    # Run locally

Links

• npm
• GitHub
• Author

License

MIT - Aman