Watch the work add up.
Count every chip.
See what your AI coding sessions are doing. Follow Claude Code and Codex activity, inspect token usage, and estimate API-equivalent costs across your projects—all on your Mac.
Follow live activity and inspect token usage and API-equivalent estimates in the real Chip Count interface. The built-in demo uses illustrative sessions and costs.
Every session. Every token. A little perspective.
See input, output, cached tokens, and model usage
Estimate API-equivalent costs with visible pricing assumptions
Compare sessions and explore trends by project and model
Set local budgets and track estimated usage against them
Keep your session index on your device without uploads or API keys
A few moments inside Chip Count.
From the first screen to the details that make it useful.
See where the tokens go
Explore the application’s usage analytics, model mix, and project trends in the light theme. Figures come from the built-in sample workspace, not personal session logs.
Browser preview · sample sessions · October 3, 2026Know what this preview does.
Chip Count reads supported Claude Code and Codex logs on your Mac. Its separate local index leaves source logs unchanged. No login, API key, or log uploads are required. Costs are API-equivalent estimates, with inferred rates and unpriced usage identified—not your subscription bill.
Take Chip Count for a spin.
Download the current Chip Count preview.
macOS
Developer ID signed · Apple notarized
Apple Silicon · M-series Macs
Download for macOSv0.1.1 · 8.9 MB · .dmgDownload app ZIP8.1 MB · .app.zipSHA-256 checksums
DMG313eb421f4a12bd43e5fec7f26617f6c25773df68c17713e1b98dd23d05fc22cApp ZIP46d80d72ac5c7672969084be94870a4c4bba03e424dba4d7f1adf163c96a9db9Open the DMG and drag the app into Applications, or unzip the app and move it there.
Windows
64-bit Windows
Build not available yetA verified installer will appear here when it is available.
Linux
64-bit Linux
Build not available yetA verified installer will appear here when it is available.
Notarized macOS preview. These macOS packages are Developer ID signed, notarized by Apple, and include stapled notarization tickets. The app and disk image passed Gatekeeper verification.
Read Apple’s guidance for opening downloaded apps before installing.
Download all checksumsWhy I built it.
AI coding generates a lot of work that is easy to lose track of. I wanted to see the sessions, the tokens, and the rough value of that work in one place. Watching the count move is part of the fun.
Before the first launch.
Which Mac download should I choose?
Choose Apple Silicon for an M-series Mac, or Intel for an Intel Mac. A universal build supports both. Only the architectures listed below are available. The DMG includes an app you can drag into Applications; the ZIP contains the same app without the disk image.
Is this a finished release?
No. Version 0.1.1 is a development preview. The macOS packages are Developer ID signed and notarized by Apple, with stapled tickets verified before publication. Notarization does not change the app’s development-preview status.
Are these my actual AI costs?
Costs are API-equivalent estimates based on observed tokens and available model rates. They do not represent ChatGPT or Claude subscription charges. Codex auto-review usage can use a rough inferred rate when the log omits a model; that assumption is labeled. Unsupported usage stays unpriced, so a partial estimate is never presented as a complete bill.
What does this app need?
Chip Count reads supported Claude Code and Codex logs on your Mac. Its separate local index leaves source logs unchanged. No login, API key, or log uploads are required. Costs are API-equivalent estimates, with inferred rates and unpriced usage identified—not your subscription bill.
Build it on your machine.
Requires Node.js, Rust, and the Tauri platform prerequisites. Claude Code or Codex session logs are needed for personal analytics; the included demo workspace lets you explore with sample sessions.
npm ci
npm run desktopLOCAL DEVELOPMENT · v0.1.1