Tally tracks both Claude and Codex usage in real time, with reset windows, live coding sessions, model names, token totals, and drag-and-drop handoffs between agents.
Refactor this component to use hooks.
Plan a 3-day trip to Lisbon.
Reply to this email for me?
Two providers, one quiet menu-bar view ↑
Choose which provider appears in the menu bar. In dual mode, each ring sits beside its own percentage. Outer arcs show usage; inner arcs show the reset window.
Click for session and weekly rings with live reset countdowns - plus a thin inner ring for your per-model Fable weekly limit. Three color themes, four looks (from dark glass to Graphite e-ink), and an optional battery-style display. Prefer bars? One toggle flips the whole view.
See Claude Code and Codex sessions with names, effective models, billed tokens, and a traffic light for what each is doing. Click to jump to its terminal tab, or switch models directly from Tally. Recent claude.ai chats sit below.
Set a threshold - say 80% - and get pinged separately for session and weekly. Surprise cut-offs, gone.
Tally uses your own Claude and Codex logins and has no account system or analytics. Claude secrets stay in Keychain; Codex owns its ChatGPT OAuth credentials. Local session files are read-only and never uploaded.
Drag a Claude Code or Codex session onto the other provider. Tally carries its local conversation and repository state across—even when the first agent has reached its usage limit.
Every feature, for everyone - no account, no cost, no catch.
Download it, use it, keep it. Every feature works for everyone - no account, no cost, no catch.
A friendly joke. After a week, Tally may gently ask whether you'd like to chip in. It never stops working and nothing is taken away.
If Tally saves you from a few surprise cut-offs, you can support the project. Entirely optional - the app is fully yours either way. 💙
Independent project, not affiliated with or endorsed by Anthropic or OpenAI. “Claude,” “ChatGPT,” and “Codex” are trademarks of their respective owners.