tokenburning

How much did you actually squeeze out of your AI subscriptions

$100 plan → $4,000 of tokens at API prices. tokenburning reads the local logs of Claude Code, Cursor, Codex, Gemini CLI, Copilot, OpenCode and Cline, prices them with live model pricing and shows your squeeze ratio (×N vs API), where the spend goes and how you actually work. Installs in seconds, sends nothing by default.

See your ×N 
curl -fsSL https://raw.githubusercontent.com/rshatskiy/tokenburning/main/install.sh | sh
tokenburning dashboard — cost, tokens, breakdowns and session analytics

Local-first

The collector runs on your machine, the dashboard binds to 127.0.0.1. Zero network egress by default.

Honest

Unpriced models flagged ~est. Session analytics labeled signal, not exact. No guessing dressed up as fact.

Private by construction

Only derived aggregates leave (if you opt in): no source, no prompts, no project paths. The server cannot see your content.

How it works

1Install and instantly see your ×N: how far your subscription beat API prices, plus spend by tool, model and project.
2Go deeper — session signals show how you actually work, not just the bill (signal, not exact).
3Optionally connect to your team — consent-gated, aggregate-only, cohort medians at 5+.

Rolled AI out to a team? The report upstairs is one link away

"They seem to use it" doesn't sell upstairs. Leadership needs proof that adoption is happening and the subscription budget pays off. The team-lead dashboard shows who's actually onboard — and who's paying for an idle seat; value extracted with ROI against the subscription budget; and who needs help. A weekly digest lands in Telegram — a ready-made report. Consent-based and aggregate-only: no prompts, no paths, no spying on individual developers. Report up without becoming a watchdog.

Create a team