thaitype Claude Usage
Aggregates token usage and calculates estimated API cost from a session's .jsonl file.
Commands
# Show total usage for a session (falls back to $AGENT_SESSION_ID if no arg)
python3 .claude/skills/claude-usage/scripts/usage_summary.py <session-id>
# Filter by date range (ISO format, date-only or datetime)
python3 .claude/skills/claude-usage/scripts/usage_summary.py <session-id> --start 2026-03-01
python3 .claude/skills/claude-usage/scripts/usage_summary.py <session-id> --start 2026-03-01T09:00:00 --end 2026-03-16T23:59:59
# Use custom projects directory
python3 .claude/skills/claude-usage/scripts/usage_summary.py <session-id> --dir /path/to/projects
Output
Session: <id>
Period: 2026-03-01 09:00 → 2026-03-16 18:42
Turns: 42
Tokens
Input: 1.2m
Cache write: 45.3k
Cache read: 890.1k
Output: 123.4k
Estimated cost: $1.2345
By model: (shown only if session used multiple models)
claude-sonnet-4-6 38 turns $1.10
claude-opus-4-6 4 turns $0.13
How cost is calculated
Each assistant turn's tokens are multiplied by the model's per-token price and summed:
| Token type | Billed as |
|---|---|
input_tokens |
input price |
cache_creation |
cache write |
cache_read |
cache read |
output_tokens |
output price |
Prices are estimates based on public Anthropic pricing. Actual billing may differ.