Favicon of Codeburn

Codeburn

A local AI coding cost tracker reads Claude Code, Cursor and Codex sessions to break down token spending by model, project and task.

Screenshot of Codeburn website

Codeburn helps developers see which AI coding work consumes their tokens and budget. It reads existing session files from tools such as Claude Code, Cursor, Codex, Copilot and Gemini CLI, so you can compare spending across agents in one place.

  • Cost breakdowns: Compare spending by model, project, branch and activity, then compare date ranges to check how a workflow change affected costs.
  • Session and PR analysis: Trace costs through individual sessions and pull request phases, including reviews and follow-up work; Claude Code analysis also identifies expensive prompts, repeated context reads and failed steps.
  • Waste detection and fixes: Find repeated file reads, oversized context and unused MCP servers, review estimated savings, and preview reversible configuration changes.
  • Usage monitoring: Track daily spending, projected monthly totals and connected providers' usage limits through desktop, editor and menu bar views.

The free, open-source app uses the MIT license and runs on macOS, Windows and Linux. It also has a CLI, a local browser dashboard and an extension for VS Code, Cursor, Windsurf, Antigravity and VSCodium. The native macOS app includes a floating dock for provider limits, and Linux has a GNOME panel extension.

Core tracking needs no account or API key. Prompts, code and project names stay on your computer, and PR analysis uses local transcripts rather than fetching data from GitHub. Pricing data refreshes online through LiteLLM. The desktop app offers optional anonymous telemetry.

Similar to Codeburn