What I'm using Claude Code for, and how much. Generated from local transcripts on 17 May 2026.
17.53Btokens through Claude Code since Mar 2026
97,258messages
41active days
31day streak
628.8Mnet new tokens
~96% of the headline is cache reads (free replay of context the model already saw). The honest "did real work" number is 628.8M in input + output + cache writes. Both numbers are accurate, the second is the one I think about.
Daily activity
Last 60 days. Each cell is one day, intensity is messages sent.
LessMore
Models
Share of input + output + cache writes by model.
Claude opus 4.782.0%
Claude haiku 4.514.2%
Claude opus 4.63.1%
Claude sonnet 4.60.7%
Tools
How Claude actually spends its turns. 55,370 tool calls across 75 distinct tools. More reads than edits usually means I'm in exploration mode.
Bash52.1%
Read15.2%
Edit12.5%
Write4.2%
TaskUpdate4.0%
TaskCreate2.3%
WebSearch1.6%
WebFetch1.3%
Grep1.2%
ToolSearch0.8%
other (65)4.7%
Meta
First entry2026-03-06
Last entry2026-05-17
Peak week4.44B tokens Β· w/o 2026-04-27
Projects scanned58
Files scanned7,621
Aggregated locally by scripts/aggregate-claude-usage.mjs from ~/.claude/projects/**/*.jsonl. No prompt content is read or shipped. Categories are curated in content/claude-categories.json.