Token Meter for Claude

See how many tokens your message and Claude's reply will likely use — before you hit send. Token Meter adds a small, draggable widget to claude.ai that estimates token usage before you hit send, so a long task doesn't run out of context partway through.

It shows:

• Your message's token count, live as you type
• A predicted reply size, as a calibrated range (not a misleading single number)
• A context-window bar — how much of the window your conversation has used, and whether your next exchange will fit
• A "plenty of space / getting close / may not fit" badge

How the prediction works:

• A small neural network predicts reply length from your prompt's wording and any size you ask for ("300 tokens", "2 pages").
• It personalizes to you: after each reply it compares its guess to the actual length and adapts, getting more accurate the more you use it.
• Reply length is genuinely variable, so it reports a calibrated range rather than pretending to know the exact number.

Privacy: everything runs and stays on your device. Your conversations are never sent to the developer or any third party. No analytics, no servers.