What is CodeBurn?
CodeBurn shows where your AI spend goes. The README explains that a bill gives a month total without breaking it down by project, model or task. Claude Code, Codex, Cursor and similar tools each write a session file whenever you use them, those files hold every token and model call, and CodeBurn reads them to produce the breakdown. There is no account and no sign-up.
Run npx codeburn to try it, or keep it with npm install -g codeburn or Homebrew. It needs Node.js 22.15 or later. Tables show cost by tool, model, project and task, and the interface is available in the terminal, in a browser, in a desktop app and as a macOS menu bar view. Compare periods puts two date ranges side by side to test whether a model or workflow change cost less.
The optimize command reads your sessions and config and lists what costs tokens without earning them. A guard feature is opt-in and local, and a local MCP server over stdio lets an agent query your usage data. The macOS builds are signed with a Developer ID and notarized by Apple.
Key capabilities
- Cost breakdowns By tool, model, project, git branch and task.
- Session drill-down From totals to sessions and turns.
- Period comparison Two date ranges side by side.
- Optimize command Lists config and habits that waste tokens.
- MCP server Local usage data available to your agent.
Where CodeBurn fits
- Finding expensive sessions and projects
- Comparing model or workflow changes over time
- Attributing cost to a feature branch
- Spotting token waste in configuration
- Limiting runaway sessions with the opt-in guard
Install
npm install -g codeburnReviews of CodeBurn
Be the first to share your experience with CodeBurn.
Write a review
Loading review form…
Alternatives to CodeBurn
CoucouCoucou displays AI coding sessions, file changes and approval requests in a desktop overlay, with companion apps for iPhone and Android. It also provides AI chat, dictation and call transcription.DDevChainDevChain orchestrates coding agents through a browser interface with shared tasks, chat, live terminals and code review. It runs on user-owned hardware and uses existing provider accounts.
MulticaMultica is a source-available workspace for assigning issues to AI coding agents, tracking runs and reviewing results, via cloud, desktop or self-hosting.