Calcaas already models what your AI product should cost. Now it tracks what your AI tools actually cost — real Claude Code token usage, by model, by tool, by device, across every machine you code on. Your prompts never leave your computer.
Project pricing and margins with the calculator; track real token spend with Usage Analytics — forecast to actuals, in one tool.
Start tracking in 60 seconds
One small companion CLI, one per-device key, and your dashboard starts filling in.
1
Install the companion
A tiny open-source CLI that wraps ccusage. It runs on your machine and reads your local Claude Code logs. Requires Node 20+.
npm i -g calcaas-sync
2
Link a device
Generate a per-device key in your dashboard, then log in from the machine. Name each one — "work-laptop", "desktop" — so usage breaks down per computer.
calcaas-sync login
3
Sync and watch
Push sends numeric daily aggregates once; watch keeps it live on an interval. Your dashboard fills in instantly — no prompt content ever leaves your machine.
calcaas-sync push
calcaas-sync watch
Everything you wish you already knew
The numbers your AI tools never surface — organized, broken down, and yours.
Every model, broken down
See which models eat your budget — input, output, and cache tokens, ranked by cost, with capability badges pulled straight from the Calcaas catalog.
Daily trends at a glance
Tokens and cost per day on an interactive timeline. Spot the afternoon a big refactor quietly cost you $40.
One dashboard, every machine
Code on a laptop and a desktop? Each device reports separately and rolls up into a single view — and a single total.
Budgets that warn you
Set a monthly spend cap — all tools or just one — with a custom alert threshold, and watch the progress bar before you blow past it. (Email alerts rolling out.)
Live or idle, at a glance
A status dot shows whether each device is actively syncing, idle, or stale — so you always know how fresh the numbers are.
Free for everyone
No paywall, no premium gate. Usage Analytics is free for every signed-in Calcaas account.
Numbers leave. Prompts don’t.
Usage Analytics is built so the sensitive part of your work never moves. The companion CLI reads ccusage’s numeric output — and only that.
Only numeric aggregates are ever transmitted — token counts, model name, date, and cost.
No prompt or response content is ever read or sent. Not once.
The sync engine is the open-source ccusage — you can read exactly what it does.
Per-device keys are stored only as SHA-256 hashes; revoke any device in one click.
Your data is row-level-isolated in Supabase — only you can read your own usage.
Honest about the numbers. Today Calcaas tracks Claude Code; Codex, Gemini CLI, Copilot, and OpenCode are coming via the same engine. Token counts come from your tools’ local logs and are estimates; costs use ccusage’s LiteLLM-based pricing. Treat them as a reliable directional read, not an invoice.
Questions, answered
No. The companion CLI only consumes numeric token aggregates from ccusage — counts, model names, dates, and cost. No prompt or response content is ever read or transmitted.
See your real AI spend
Free for every account. Sign in, generate a device key, and watch your dashboard fill in.