You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
-[Claude Code](https://docs.anthropic.com/en/docs/claude-code) installed and used (generates the local data TokenMeter reads)
62
+
-[Claude Code](https://docs.anthropic.com/en/docs/claude-code) installed and used
62
63
63
64
## How It Works
64
65
65
-
TokenMeter reads Claude Code's local JSONL files from `~/.claude/projects/`:
66
+
### Rate Limits (real data)
67
+
TokenMeter reads Claude Code's OAuth token from the macOS Keychain and calls `api.anthropic.com/api/oauth/usage` to get real utilization percentages and reset times. On first launch, macOS will ask you to allow Keychain access — click "Always Allow".
66
68
67
-
1.**Daily usage** — scans all conversation JSONL files, groups by date, calculates token costs using embedded model pricing
68
-
2.**Rate limits** — tracks output tokens in the last 5 hours (session) and 7 days (weekly), shows progress against estimated plan limits
69
-
3.**Model breakdown** — identifies which models (Opus, Sonnet, Haiku) are being used and their relative costs
69
+
### Usage Analytics (local data)
70
+
Parses JSONL files from `~/.claude/projects/` and `~/.config/claude/projects/`:
71
+
-**Daily costs** — groups by date, calculates using embedded model pricing
72
+
-**Hourly heatmap** — groups by hour-of-day per date for activity patterns
73
+
-**Model breakdown** — per-model token and cost breakdown
74
+
- Deduplicates by `requestId` and filters `<synthetic>` entries
70
75
71
-
No API keys, no external tools, no cloud services. Everything is local.
76
+
### Notifications
77
+
Sends macOS notifications when rate limit utilization reaches 80% (session or weekly). Throttled to once per hour per window. Toggle in Settings.
72
78
73
79
### Pricing
74
80
75
-
Costs are calculated using Anthropic's published API-equivalent pricing (per million tokens):
81
+
Costs are calculated using API-equivalent pricing (per million tokens):
0 commit comments