A tiny meter in your menu bar that shows, live, how much your AI coding tools are costing you — your token count, your dollar spend, and how close you are to your plan limit. It reads the logs your tools already write on your own computer, so there are no API keys and nothing to configure. It also feeds those numbers to your phone and watch.
The desktop meter is the same live monitor the paid apps build on. It reads the logs your AI coding tools already write on your computer and turns them into a running total.
Counts today's tokens and dollar cost across Claude desktop, Claude Code, Codex, Gemini CLI, and more — all in one number.
Shows your rolling 5-hour window and weekly limit with reset times, so you wrap up before you get cut off.
Pushes the same numbers to the Vibe Guard app through your account — check your usage anywhere.
No API keys. It only reads usage counts from local logs — never your prompts or your code.
Same live meter in the corner of your screen — whether it's your first app or your fiftieth.
You've no idea what an AI session actually costs. (A token is the unit of text an AI reads and writes — what you pay for.) The meter shows a plain dollar figure ticking up as you build — so you stop before the surprise, not after it.
You run Claude Code, Codex, Cursor, and Gemini CLI in parallel. The meter rolls all of them into one running total and counts down your 5-hour and weekly limits, so you wrap up before you get cut off.
After Microsoft pulled engineers off Claude Code when AI spend blew the budget, per-seat visibility stopped being optional. The meter reads real usage locally — no provider API, no keys to hand out.
The first time it runs, sign in with your Vibe Guard account — that links this computer to you automatically. No codes to copy.
./start-vibeguard-meter.sh (Python 3 is usually built in).Start with the meter. Upgrade later for all-tools aggregation, the context-window gauge, limit alerts, and the on-device Coach.