Tally — Claude token counter

Live token counter for Claude. Model-aware messages-left, reset countdowns, and one-click handoff to ChatGPT/Gemini/Grok.

Tally is a quiet meter for Claude. It lives where you work: a thin strip beneath claude.ai's message box that shows your real usage in real time, predicts how many messages you have left, and lets you continue the conversation in ChatGPT, Gemini, or Grok in one click. WHY TALLY Claude's own settings page tells you "Current session: 31%", but not how many messages 31% buys you, and not what happens when you hit the 5-hour limit. Tally fixes both: • Live inline meter: 31% used, about 12 opus messages, resets in 1h 22m • Model-aware estimates: Opus burns roughly 5x faster than Sonnet and 25x faster than Haiku. The number always matches the model you are actually using. • Cross-LLM handoff: when you run out of Claude headroom, send the conversation to ChatGPT, Gemini, or Grok. Tally opens the tab, pastes your recent turns straight into the box, and automatically attaches the full conversation as a Markdown file so nothing gets truncated, including the files and code Claude built. THREE SURFACES, ONE NUMBER 1. Inline bar beneath the chat input on claude.ai 2. Sidebar card on the left nav showing session and weekly windows 3. Toolbar popup with a full breakdown matching claude.ai/settings/usage PRO FEATURES Free: 2 cross-LLM handoffs per day (unlimited copy and download). Pro: $2/mo or $20 lifetime via Polar.sh. Unlimited handoffs, a license key, and it works on every device you sign in from. PRIVACY Tally reads your Claude usage from claude.ai's own API using your existing browser session. No API keys, no new accounts, no servers. All counting happens locally in your browser. When you hand off a conversation, it stays on your machine: the recent turns are pasted and the Markdown transcript is attached locally, right in your browser. Conversation content is never sent to a Tally server, because there isn't one. NOT AFFILIATED WITH ANTHROPIC.