Model Usage — Hermes Desktop plugin
Claude / Codex subscription usage at a glance in the Hermes Desktop status bar.

- One triple-ring gauge per provider: outer thin ring = context window of the current chat, middle = 5-hour session window, inner = weekly window, provider letter in the middle (
CClaude,XCodex). - Limit alerts: when a 5h or weekly window drops to 30% and again to 10% remaining, one in-app toast plus one Windows/OS notification (shown only while Hermes is not focused). Each threshold fires once per window period.
- Color by remaining share: >50% green · 31–50% yellow · 11–30% orange · ≤10% red.
- Click → popover with every window as a bar plus reset countdown. Duplicate pool rows of the same login are merged.
- Gemini: link to Google AI Studio usage (Hermes has no Gemini usage API).
- Read-only: uses the gateway
model.optionsusage cache; never touches credentials. Refreshes every minute. - UI strings: English, 한국어, 日本語.
Install
hermes plugins install Mirinai307/hermes-plugin-model-usage
Then open Capabilities → Plugins in Hermes Desktop and switch Model Usage on (desktop halves install disabled).
License
MIT
Disclosure
- No network calls of its own: usage comes from the Hermes gateway (
model.options, the same cache the model picker uses) and the focused chat's usage stats. - Never reads, stores or sends credentials. No shell commands, no background processes, no telemetry.
- Stores only the list of alerts already shown (plugin storage) so each threshold fires once per window.
- The Gemini row opens
https://aistudio.google.com/usagein your browser when clicked. - Works on Windows, macOS and Linux (pure desktop SDK, no Python).