Where does your Claude Code money actually go?

Open-source forensics for AI coding-agent spend.

spend breakdown · my last 102 sessions $1,540 total

necessary spend$1,340

late compaction$123

oversized tool results$61.65

redundant re-reads$13.73

stale continuation$1.83

avoidable $200 — 13% of spend

bash

$ git clone https://github.com/TarekAwwad/authrty-claude-code-analytics
$ cd authrty-claude-code-analytics && docker compose up --build
# then open http://localhost:5173

what it finds

Spend mindmap showing habits and anti-patterns across Claude Code sessions

01 / 06 Habit & anti-pattern mindmap

Subgroup analysis of tool call errors — what combinations drive cost

02 / 06 Tool error subgroup analysis

Late compaction forensics — context window cost breakdown

03 / 06 Late compaction forensics

Detailed token spending across Claude Code sessions

04 / 06 Project & session cost

05 / 06 Session forensics

06 / 06 Turn outlier analysis