Changelog

All notable changes to TokenGauge will be documented here. This project adheres to Keep a Changelog and Semantic Versioning.

0.0.1

First release line: a native-only, privacy-first multi-agent gauge cockpit.

Native multi-agent cockpit. Claude Code and Codex appear as first-class
per-agent gauge cards with plain-language provenance and freshness badges.
Raw source, freshness, and accuracy metadata is available in Diagnostics.
Claude native snapshots. Reads a passive local statusLine snapshot your
own statusLine writer produces (opt-in bridge) plus per-model cost and model
information from the local stats-cache.json cache.
Codex native app-server probe. On explicit opt-in (off by default), asks
the local codex app-server for account rate-limit information. This version
recognizes the tested 5-hour and 7-day account-window shape; different bucket
shapes show unavailable/unsupported instead of guessed values. Nothing is
spawned while it is off or while the Codex card is hidden.
Honest states. Missing native data reads unknown/unavailable; missing cost
reads cost unknown. TokenGauge never reconstructs, estimates, or synthesizes
values it cannot read natively.
Secret handling, diagnostics, and Command Palette workflows. TokenGauge
stores no API keys or provider credentials; the local install salt lives only
in VS Code SecretStorage via SecretManager; redacted cockpit diagnostics and
a privacy report; release-docs and release-workflow safety gates (release
automation itself is added at release time).

Posture for this release:

No developer-controlled telemetry; no outbound network by default.
GitHub Release first distribution is planned; Marketplace and Open VSX publishing are optional PAT-gated paths.
Every metric is accuracy-labeled; native-reported values are never presented as provider billing data.
TokenGauge stores no API keys or provider credentials; the local install salt lives only in VS Code SecretStorage via SecretManager.

Internal cleanup (native-only reset):