Quickstart

From zero to your first cost breakdown in under a minute - nothing added to your stack.

Run it

The fastest way to try CodeBurn is with npx. It runs locally, reads logs your tools already write, and changes nothing about how they run. No config, no API keys:

npx codeburn

CodeBurn auto-detects your AI coding tools, reads their session data locally, prices every call with LiteLLM, and opens the interactive dashboard for the last 7 days. Everything runs on your machine.

Before you begin

Make sure you have:

• Node.js 22.13 or newer
• At least one supported AI coding tool with session data on disk - Claude Code, Codex, Cursor, Copilot, Gemini, and 30+ more

Step by step

1. See your usage

npx codeburn

In the dashboard, press 1–6 to switch period (Today, 7d, 30d, Month, 6m, Lifetime), p to toggle provider, and q to quit.

2. Break it down by model

npx codeburn models

Per-model token and cost table, optionally exploded by task type.

3. Find waste

npx codeburn optimize

Scans for token-waste patterns and gives copy-paste fixes.

4. Install it (optional)

Prefer a plain codeburn command without npx?

npm install -g codeburn

5. Menu bar app (macOS)

Keep today's spend in your menu bar:

codeburn menubar

What's next

• Dashboard: read the interactive view
• Supported tools: everything CodeBurn tracks
• Optimize: cut token waste
• Menu bar: the native macOS app