WhereMyTokens — Windows tray app for real-time Claude Code token monitoring

[jeongwookie]

WhereMyTokens

Windows system tray app that shows Claude Code token usage, costs, and rate limits in real time.

https://github.com/jeongwookie/WhereMyTokens

---

What it does

Sits quietly in the taskbar and surfaces everything you need to know about your Claude Code usage at a glance:

• Live session tracking — detects active Claude Code sessions across Terminal, VS Code, Cursor, Windsurf, etc. with real-time status (active / waiting / idle / compacting)
• Rate limit bars — 5h and 1w usage from the Anthropic API, with progress bars and time-to-reset counters
• Claude Code bridge — registers as a statusLine plugin to receive live rate-limit data from Claude Code via stdin (no polling)
• Context window warnings — per-session compact warnings at 50% / 80% / 95%+ inline in the session list
• Tool usage bars — color bar + top-3 tool breakdown (Bash, Edit, Read…) per session
• Activity heatmaps — 7-day heatmap and 5-month GitHub-style calendar grid
• Model breakdown — per-model token and cost totals
• Alerts — Windows toast notifications at configurable thresholds
• Always-on-top widget — global hotkey to toggle, minimize to tray
• Tray label — show usage %, token count, or cost directly in the taskbar

Why I built it

I kept losing track of how many tokens I was burning across sessions and hitting rate limits unexpectedly. I wanted something passive — always visible in the corner, no browser tabs, no manual refreshing. So I built a tray app that reads Claude Code's local JSONL files directly and supplements with the Anthropic API.

Platform

Windows only (Electron + React). Pre-built portable .exe available — no install needed.

Would love feedback, especially on the bridge integration or any feature requests!

[affaan-m]

This is useful. The statusLine bridge and local JSONL reading are exactly the right integration points for a passive Windows monitor.

The two things I would be careful about for ECC users:

1. Privacy: make it very clear which local session files are read and what, if anything, leaves the machine.
2. Rate-limit/API access: keep Anthropic API credentials out of project repos and document the storage path/permissions for any key material.

If you want this to be easier for people here to evaluate, a short architecture/security section in the README would help: data sources, local storage, network calls, credential handling, and how to uninstall/disable the statusLine bridge.

I am glad you shared it. Token/cost visibility is one of the highest-leverage missing pieces in long Claude Code sessions, especially on Windows.

[jeongwookie]

@affaan-m
Thanks for the careful review. I updated the README with dedicated Architecture and Security & Privacy sections covering local data sources, app storage paths, network calls, credential handling, rate-limit/API backoff, and how to disable or remove the Claude Code statusLine bridge.

I also added a Settings -> Claude Code Integration -> Disable action. It only removes the WhereMyTokens-owned statusLine entry and leaves any custom statusLine unchanged.

The app remains local-first: no telemetry or cloud sync. Local session logs stay on the machine, and provider credentials are read from the official local CLI files only for usage requests and token refresh 👍

[affaan-m]

Thanks for the update. The added architecture, security, privacy, credential-handling, backoff, and uninstall/disable notes are the right review surface for ECC users.

The statusLine ownership rule matters: removing only the WhereMyTokens-owned entry and leaving custom statusLine behavior untouched is the kind of integration boundary I would want. Local-first with no telemetry/cloud sync is also the right default for session/token visibility. Closing this thread as acknowledged and triaged; please open a new discussion or PR if the integration surface changes materially.

[affaan-m]

Love this — a Windows tray app for token monitoring is a legit missing piece in the Claude Code ecosystem on Windows — the platform-native experience there is still rough.

A few things from our side:

1. Sponsor tier refresh just went live — if ECC saves you time enough that you'd like your name in the wall of sponsors, the Builder tier is $25/mo (was $10) and gets you listed in SPONSORS.md plus a private monthly progress note. No pressure on the pitch — just making it easy to find.

2. Happy to feature this in a Show & Tell roundup on our next release if you're up for it — I'd link your repo/gist/post in the release notes and tweet from @affaanmustafa. DM works or reply here.

3. Keep building.

— Affaan

[Amar07Singh]

Hey @jeongwookie, curious how Claude Code is handling your ECC setup so far?

Are you hitting those weird context walls where it starts losing track of files, or is it staying snappy?

[jeongwookie]

@Amar07Singh So far it has been pretty snappy for my ECC + Claude Code workflow.

I haven't hit many hard "lost the whole project" context-wall moments during normal scoped tasks. Claude Code usually handles the repo fine as long as the task is bounded and I point it at the relevant files.

Where it can still drift is in longer sessions that touch multiple layers at once, like Electron main/preload/renderer plus docs, release notes, and privacy/security details. After compaction or a very long thread, I usually re-anchor it with AGENTS.md, explicit file paths, and a quick build/privacy check.

So my take is: ECC doesn't remove context limits, but it makes the session state much more visible and easier to manage. For my setup it has stayed usable and fast, especially when I keep the work split into focused chunks.