Tracks multiple AI coding tools
CodeBurn reports usage across 40 tools, including Claude Code, Cursor, Codex, Copilot, and Gemini, so developers can compare spend across different assistants in one place.
CodeBurn
BeanspruchenCodeBurn is a local-first tracker for AI coding token usage and cost across tools like Claude Code, Cursor, Codex, Copilot, and Gemini. It helps developers see spend by model, project, task, and session without sending their data off device.
Overview
CodeBurn is a local-first AI coding cost and token usage tracker for developers who want to understand where their usage spend goes across multiple tools. It reads session data from local files, prices calls using LiteLLM, and turns that data into breakdowns by model, project, task, activity, and time period.
The product is available as a terminal command, a desktop app, and a macOS menu bar view, with a GNOME extension for Linux. It is designed to work without API keys, wrappers, or proxies, and it keeps processing on the device by reading local logs only.
Features
Breaks spend down by workflow dimensions
Reports can be grouped by model, project, task, activity, session, and tool, making it easier to see what work consumed the budget.
Runs fully locally
The app reads session files already on disk and does not require uploaded logs, API keys, or a proxy layer.
Offers multiple interfaces
Users can check usage from the terminal with npx codeburn, from a macOS menu bar app, or from a desktop dashboard.
Includes efficiency-oriented views
Documentation and site copy mention per-model tables, one-shot rate, cache hits, retry tax, and optimization views for spotting waste patterns.
Supports time-based analysis
Users can switch between today, week, month, six months, or lifetime views, and see projections such as month-to-date spend and expected month total.
Use Cases
Monitoring day-to-day AI spend
A developer using one or more coding assistants can keep the current day’s cost visible in the menu bar or terminal instead of waiting for an end-of-month bill.
Comparing tools and models
Someone switching between Claude Code, Cursor, Codex, or other supported tools can review which model or provider is consuming the most budget.
Reviewing where work time went
Teams or individual developers can look at breakdowns by coding, debugging, feature work, and related activities to understand what kinds of tasks are most expensive.
Finding inefficient sessions
The Optimize and compare workflows can help users identify patterns such as retry-heavy work, low one-shot rates, or expensive sessions that may need adjustment.
Connecting spend to shipped work
The Yield view and session breakdowns are aimed at users who want to relate AI usage to git commits, branches, or pull request activity.
Pros and Cons
Pros
- Works locally and is described as privacy-first, with no data upload required for core tracking.
- Covers a broad set of AI coding tools rather than a single provider.
- Offers both terminal and visual interfaces for different workflows.
- Breaks usage down into practical dimensions such as model, project, task, and session.
- Includes optimization-oriented views, not just raw totals.
Cons
- The site does not provide clear pricing or plan details, so cost of the product itself is not evidenced here.
- Some integration depth is unclear from the available pages, beyond reading local session files and auto-detecting tools.
- Team and multi-user collaboration workflows are not fully documented in the source provided.
FAQ
How does CodeBurn get usage data?
It reads local session files already stored on the machine, such as JSONL files, SQLite databases, or JSON, and then prices the recorded token usage.
Does CodeBurn require API keys or a proxy?
No. The documentation says it runs locally and does not need API keys, wrappers, or a proxy.
Which platforms does it support?
The site and comparison page describe support for macOS, Linux, and Windows, with a macOS menu bar app and a GNOME extension on Linux.
Can it track more than one AI coding tool?
Yes. The homepage and docs describe support for 40 tools, and the comparison page frames it as useful for people using more than one assistant.
Is there a paid plan listed on the site?
No plan or pricing details were available from the source pages provided, and the pricing URL returned a 404.
Quick Facts
- Category
- Developer Tool
- Primary use
- AI coding cost and token usage tracking
- Platforms
- macOS, Linux, Windows
- Deployment model
- Local-first, reads session data from disk
- Source domain
- codeburn.app
- Pricing evidence
- No pricing page found in the provided source set