Add Intel Mac support with x64 DMG, updater manifest, and Homebrew cask routing.

Add Claude Fable 5 pricing (10ドル/50ドル per MTok) so Fable usage is costed correctly instead of falling back to generic Claude rates.

Tokcat 0.1.29

• Fix menubar popover behavior across macOS spaces.

• Count complete final JSONL usage lines even without a trailing newline.

Tokcat 0.1.28

Keyboard shortcuts:

• Global Ctrl+Cmd+T toggles the popover from anywhere
• In-popover: Cmd+, settings, Cmd+R refresh, Cmd+W/Esc close, Cmd+Q quit, Cmd+1-9 jump to tab, Cmd+[ / Cmd+] prev/next tab, Cmd+G 2D/3D, Cmd+U updates
• Hold Cmd to reveal shortcut hints on each control

Polish:

• Refresh icon spins while a manual refresh is in flight
• Dismissing the popover returns focus to the previously active app
• Cmd+W / Esc closes an open settings/about panel before hiding the popover

Tokcat 0.1.27 — pricing estimates refresh

Update bundled model pricing for current Claude Opus, GPT-5/o-series, Gemini, GLM/Kimi, and Cursor Composer models. Add tiered pricing support for long-context estimates.

Tokcat 0.1.26 — center-aligned usage stats

Center-align the Total / Tokens / Best day columns in the combined usage card.

Tokcat 0.1.25 — 2D/3D usage toggle

Toggle the dashboard's usage view between a 2D stacked-by-agent bar chart
and the 3D full-year contribution calendar. Token Usage stats now lead the
card and stay visible in both views; switching never reflows the layout.

Add agent OAuth usage dashboard

• Add overview and per-agent dashboard tabs

• Add Codex and Claude OAuth quota usage

• Replace 3D view with 2D bar charts and hover details

• Show live session usage in overview

Tokcat 0.1.23

• Prevent tray animation stutter by disabling implicit Core Animation layer actions during frame swaps and refresh hops.
• Refresh README for live tokens/min, Live trace, and current tray animation settings.
• Include the redesigned GitHub Pages landing page.

Tokcat 0.1.22

Release dependency-free usage graph fixes.

• Align bundled pricing for current Claude, GPT-5, and OpenRouter model families.

• Use stateful Codex token parsing to avoid stale total snapshot and fork baseline over-counting.

• Preserve responsive popover height so bottom sessions remain visible.