Skip to content

Releases: sebasrodriguez/tokentap

v0.3.1 — Color-coded warnings + bug fixes

20 Mar 19:07

Choose a tag to compare

Added

  • Menu bar indicator changes color based on usage thresholds: blue (normal), amber (warning, default 80%), red (critical, default 90%)
  • App icon

Fixed

  • Notifications now actually fire when thresholds are crossed
  • Changing refresh interval in Settings takes effect immediately
  • Removed 1,000+ lines of dead code (old ClaudeUsageBar)

Full changelog: v0.3.0...v0.3.1

v0.3.0 — Pure API: No more CLI spawning

20 Mar 15:27

Choose a tag to compare

Highlights

Both providers now use direct API calls instead of spawning CLI processes. This makes polling dramatically faster, more reliable, and simpler.

Claude Code — New API-based provider

  • Calls api.anthropic.com/api/oauth/usage directly
  • Reads OAuth token from macOS Keychain (no CLI binary needed)
  • Returns structured JSON: session %, weekly %, Sonnet %, Opus %, reset times
  • No more PTY spawning, ANSI parsing, or regex extraction

OpenAI Codex — Unchanged

  • Still reads rate limit headers from chatgpt.com/backend-api/codex/responses
  • Uses gpt-5-codex-mini for minimal usage impact

What changed

  • 160 lines added, 312 removed — the codebase got simpler
  • Shared PollerError type across providers
  • Claude settings: removed CLI path (no longer needed)
  • Added Opus weekly usage tier for Claude

Install

brew tap sebasrodriguez/tap
brew install tokentap

Full changelog: v0.2.1...v0.3.0

v0.2.1 — Fix Codex provider

20 Mar 13:20

Choose a tag to compare

Fixed

Codex provider completely rewritten. The PTY-based approach couldn't capture Codex's /status output (rendered in an alternate screen buffer).

New approach: Makes a minimal API call to chatgpt.com/backend-api/codex/responses and reads rate limit data directly from response headers:

  • x-codex-primary-used-percent — 5h session usage
  • x-codex-secondary-used-percent — weekly usage
  • x-codex-primary-reset-at / x-codex-secondary-reset-at — reset timestamps

Uses gpt-5-codex-mini (cheapest available model) to minimize usage impact per poll.

Full changelog: v0.2.0...v0.2.1

v0.2.0 — TokenTap: Multi-Provider AI Usage Tracker

19 Mar 21:07

Choose a tag to compare

What's New

TokenTap (formerly Claudometer) now supports multiple AI providers with a pluggable architecture.

Multi-Provider Architecture

  • Claude Code — session, weekly, and Sonnet usage with reset times
  • OpenAI Codex — usage percentage with model info
  • Menu bar auto-rotates between providers (configurable: 10/20/30s)
  • Dropdown shows all providers at once with their usage tiers
  • Per-provider settings tabs for refresh interval and CLI path

Improvements

  • Parser rewritten for reliability (handles single-line TUI output)
  • Cached data preserved on poll failure
  • Menu bar updates reliably via objectWillChange forwarding

Install

brew tap sebasrodriguez/tap
brew install tokentap
open $(brew --prefix)/opt/tokentap/TokenTap.app

Or build from source:

git clone https://github.com/sebasrodriguez/claudometer.git
cd claudometer/TokenTap
make install
open TokenTap.app

Full changelog: v0.1.0...v0.2.0