tare shows where Claude Code spends your tokens

Kelviq released tare — an open-source skill (MIT) for Claude Code that reads local session logs and calculates actual token usage by project, model, file, MCP server, and subagent, compiling the results into an HTML report.

Compiled by Sergey KostenchukPublished 2026-08-28Updated 2026-09-18

2026-08-28 Coding Anthropic

🛠 tare shows where Claude Code spends your tokens

tare is out — an open-source skill (MIT) from Kelviq: it reads local Claude Code session logs and calculates actual token usage by project, model, file, MCP server, and subagent.

🌍 Claude Max plans hit limits, while hidden consumers — background scripts, MCP servers, short sessions — were invisible. At Kelviq: 1,555 hidden sessions in a day (9,022 requests) versus 93 human requests. tare also counts “amplified tokens”: a read file is resent in every subsequent message, and the cost is attributed to the tool that invoked it.

👤 Hit a limit — install the skill with one command and ask “where did my tokens go this week”: you’ll get the culprit, not a wall of numbers. /tare share provides an anonymized summary without prompts or paths. Everything is local; requires macOS/Linux and Python 3.9+.

Source 1: https://github.com/kelviq/tare Source 2: https://www.kelviq.com/blog/claude-code-usage-limits-where-tokens-go/

Sources