claudepeek
v1.0.4
Published
Local dashboard for visualising Claude Code token usage and estimated costs. Built with Claude Code (AI-generated).
Maintainers
Readme
claudepeek
A local dashboard for visualising your Claude Code token usage and estimated costs.
npx claudepeekNote: All cost figures are estimates calculated from token counts × Anthropic list prices. They are not your actual invoice. For billing, check console.anthropic.com.
Requirements
- Node.js v18 or later
- Claude Code installed and used at least once (so local usage files exist)
Usage
No install — run directly:
npx claudepeekOr install globally:
npm install -g claudepeek
claudepeekOpens http://localhost:3737 automatically.
Options:
--port <number> Use a different port (default: 3737)
--no-open Don't open the browser automaticallyWhat it shows
| Tab | What you see | |---|---| | Overview | Summary cards, daily spend chart (stacked by model), model cost breakdown, top 10 most expensive days | | Daily | Full day-by-day breakdown with token type columns | | Sessions | All-time cost per project folder (not date-filtered) | | Monthly | Month-by-month summary cards and chart | | Budget | Progress against a monthly budget, burn rate, projections, and scenarios | | About | Explanation of how the tool works and what the numbers mean |
Filtering by date
Use the preset buttons in the header (Today, Last 7 days, This month, etc.) or enter a custom date range. The filter applies to Overview, Daily, and Monthly. Sessions always show all-time totals.
Budget
Set your monthly budget amount and reset day in the Budget tab. Settings are saved in your browser's localStorage and persist across visits.
How it works
Reads ~/.claude/projects/**/*.jsonl — the conversation files Claude Code writes locally — via npx ccusage. A tiny Node.js server (no dependencies) serves the data as a local JSON API. Nothing leaves your machine.
Running from source
git clone https://github.com/martinpenabad/claudepeek
cd claudepeek
node server.jsOr double-click start.command (macOS) / start.bat (Windows).
This project was built with Claude Code (AI-generated). The source code was written through a conversation with Claude and reviewed by a human.
