What TokenTracker does
TokenTracker shows developers real token use and estimated cost across many AI coding tools, locally by default, without ever capturing prompts.
TokenTracker is a local-first AI token usage and cost tracker for people who ship software with AI coding tools. It helps you see how many tokens Claude Code, Codex, Cursor, Gemini, GitHub Copilot, and dozens of other tools actually consume, and what that usage costs over time.
It is privacy-first by design. TokenTracker records token counts, timestamps, and cost estimates only. It does not read, store, or upload prompts, model responses, or your source code. The default path keeps data on your machine.
You can use it as a cross-platform CLI (`npx tokentracker-cli`), a native macOS menu bar app with desktop widgets and a pixel pet driven by real coding activity, a Windows system tray app, and a local dashboard. The same underlying usage data powers achievements, heatmaps, model breakdowns, and an optional public leaderboard when you choose to share.
The local open-source edition is free under the MIT license. Optional Cloud sync is available if you want multi-device history without changing the privacy model of what is collected. Cloud pricing is published on the site (monthly and annual plans).
TokenTracker is for developers, indie hackers, and teams who want an honest meter for AI coding spend instead of guessing from invoices. It is not a coding agent, not a prompt marketplace, and not a cloud IDE. If you need an agent framework or a hosted chat UI, choose those tools instead and use TokenTracker alongside them to measure cost.
Getting started is usually one command or a brew/cask install for macOS. Connect the tools you already use; TokenTracker discovers local usage signals without API keys for prompts. Export and deletion stay under your control on the local path.
Compared with spreadsheet tracking or provider dashboards alone, TokenTracker unifies many AI coding tools in one place, stays local by default, and never asks you to paste prompts to get a usage chart.
Best for
Developers and indie hackers who use AI coding tools and want local, privacy-first usage and cost visibility.
Not for
People looking for an AI coding agent, prompt marketplace, or hosted IDE—TokenTracker measures usage; it does not write code for you.
Questions, answered by the maker
What does it do?
It tracks how many tokens your AI coding tools use and estimates cost across Claude Code, Codex, Cursor, Gemini, Copilot, and more—without reading prompts.
Who is it for?
Developers and indie hackers who want a local, privacy-first meter for AI coding usage and spend.
How much does it cost, and is there a free plan?
Local use is free and open source under MIT. Optional Cloud sync is available; published prices are about $4.99/month or $39.99/year at checkout.
How is it different from similar products?
It is local-first, unifies many AI coding tools, and records token counts only—never prompts or source code.
Where is my data stored, and can I export or delete it?
By default data stays on your machine. You control local files; optional Cloud is only if you enable sync. You can remove local data anytime.
How do I get started?
Run `npx tokentracker-cli`, or install the macOS app with Homebrew cask, or download the Windows installer from GitHub Releases.
Is it open source, and under which license?
Yes. The project is MIT-licensed on GitHub at xiufengsun/TokenTracker.
Can I self-host it?
Yes for local/self-run use: CLI and apps run on your devices. Cloud is optional and separate.