Tokenomics

Mint for your Claude Code tokens.

A local-first dashboard that reads your Claude Code session transcripts and reframes AI spend as personal finance: where tokens go, how much of it was productive, which habits bleed money, and what to change.

Privacy

There is no backend, and no upload. This is stronger than "local-only" — there is no server that could receive a transcript. The app is a static SPA; all parsing and analysis run client-side in a Web Worker. The only optional outbound call is an advisor blurb, gated behind an explicit opt-in flag, which sends computed aggregates only — never prompt text or file paths.

Run locally

cd web
npm install
npm run dev

Opens on http://localhost:5173. The demo view loads a bundled, anonymized snapshot (web/public/demo.json) so you see real numbers immediately. Use the folder picker to point at your own ~/.claude/projects and analyze your own transcripts — nothing leaves the browser.

Scan your own machine

node tools/scan.mjs

Generates a local snapshot from your own ~/.claude/projects corpus that you can drag into the app's snapshot importer, without going through the folder picker.

Build

cd web
npm run build

Static output goes to web/dist.

Deploy

The project deploys to Vercel as a static build from web/, configured in vercel.json (build command, output directory, SPA rewrite, and security headers).

To deploy from a local machine:

npx vercel login      # interactive browser login, run this yourself
npx vercel link       # link this directory to a Vercel project
npx vercel --prod      # deploy

vercel login requires an interactive browser session, so it can't be run non-interactively — run it yourself (in this Claude Code session, prefix with !, e.g. !npx vercel login). Once authenticated, subsequent deploys can be automated.

CI (.github/workflows/ci.yml) runs a typecheck and build on every push.