Krisvid is named after my two little boys, Kristóf and Dávid. It does not mean much to them yet, but their name sits in my tray every working day. 😎
The app itself was born out of an everyday annoyance. I use three AI coding agents (because I do not pay for the largest plans), and each one kept its remaining quota somewhere else. Then it struck me: if I can pull a given model's usage out of an API or a local log, it would be rather nice to show it all in one place. I wrote it for myself first. Then it occurred to me that if I had missed this, others probably did too: so it became a free, open-source tray app that shows limits and estimated spend in one place.
Supported providers:
Claude Code – five-hour session and weekly limits, daily/weekly cost and token statistics, per-model breakdown
OpenAI Codex – limits and token counts from local session files
JetBrains AI (AI Assistant / Junie) – monthly AI credit quota and its reset time
Features:
Green / amber / red tray indicator that shows where you stand without a click
Compact panel with gauges, a seven-day cost chart and a per-model breakdown
Combined view across installed tools, with a color per provider
Five-language UI: Hungarian, English, German, French and Spanish
Resilient polling: a failed or rate-limited request keeps the last known data visible and backs off automatically on HTTP 429
Implementation:
Electron main process with a separate renderer panel and
contextIsolationenabled: the UI reaches the main process only through a narrow, predefined APIEvery provider implements the same simple interface, so adding a new one is a single folder plus one line in the registry
Codex and JetBrains AI data is read from local files only, with no network calls
Claude Code limits involve a single outbound request to Anthropic's usage endpoint, using the OAuth token already stored on the machine; cost statistics are still assembled locally from incremental reads of the transcripts under
~/.claude/projectsInstallers for Windows, macOS and Linux (exe, dmg, zip, AppImage, deb) – the macOS and Linux builds are untested so far, so if you try one and report back, I would be grateful
Data handling: no telemetry, no usage collection, no analytics. What is on the machine stays on the machine – and because the source is open, that can be verified rather than trusted.
The project is not affiliated with Anthropic, OpenAI or JetBrains. Anthropic's usage endpoint is not a documented public API and the JetBrains quota file format is internal, so the displayed values are best estimates from the available data, not official billing.
I hope you find the app useful – if you do, give it a star on GitHub!
