Token Usage Tally Token Usage Tally
Know what Claude Code really costs.
A native macOS app that reads your Claude Code session files and turns them into interactive charts, cost estimates, and usage history — all without sending a single byte off your machine.
Requires macOS 15 Sequoia or later
Your usage, visualized.
Stop guessing what Claude Code costs. Token Usage Tally parses your local session files and presents clear, interactive charts.
Interactive Dashboard
Overview stats, daily cost charts, and model breakdown pie charts — all at a glance. Click into any slice or bar to drill down.
Daily & Monthly Views
Stacked bar charts by token type, model, and project. Hover for detailed tooltips, switch between daily and monthly summaries.
Cost Estimation
Pulls model pricing from LiteLLM's public pricing data. See estimated costs per session, per day, per model — with cache savings broken out.
Session Browser
Browse every Claude Code session with a sortable, searchable table. See token counts, models used, and associated projects for each session.
Project Breakdown
See usage and costs grouped by project. Understand which codebases are driving your token spend and track subagent usage separately.
Export & Import
Back up your usage database or move it between machines. Your data is yours — export it anytime as a portable file.
How it works.
Three steps. No accounts, no configuration files, no API keys.
Grant folder access
Point Token Usage Tally at your ~/.claude folder. A one-time macOS permission prompt — that's it.
Sessions are scanned
The app reads JSONL session files, extracts token counts and model identifiers, and stores summaries in a local SQLite database.
See your usage
Interactive charts and tables show daily costs, model breakdowns, project usage, and session-level detail — updated on every launch.
Nothing leaves your Mac.
Token Usage Tally reads local files and stores data in a local database. The only network request it makes is to fetch publicly available model pricing data from GitHub. No accounts, no analytics, no tracking, no telemetry — ever. Your conversations are never read; only numeric token counts and model identifiers are parsed.
Read our Privacy Policy →