Token Forecast for Claude

Estimates how much of your Claude usage limit a message will use before you send it, and suggests a cheaper model when one fits.

Know what a message will cost before you send it. Token Forecast shows, right next to the claude.ai message box, how much of your 5-hour and weekly Claude usage limit a message will use, before you press send. It also tells you which model and effort level are enough, with the reasons, and learns your plan over time. BEFORE YOU SEND • Forecast with an 80% range, based on your prompt, attachments (images, PDF, Word, Excel, text), chat history, model, effort and cache state. • Reasoned model suggestion: which model is enough, what it looked at and how much you would save. If it is not sure, it never suggests a lighter model. • Effort suggestion from Low to Max, with the savings when your setting is higher than needed. • Topic recognition (16 areas). For health, law and money questions it never suggests a light model. • Cost breakdown, cache countdown and peak-hour countdown. WHILE YOU WORK • "At this pace, when does it run out?" for your 5-hour limit. • Hand-off summary prompt for long chats. • Threshold notifications (50/75/90%), limit-reset notification and a toolbar badge. AFTERWARDS • "Why was that so expensive?" explanations. • Weekly report: messages, limit spent, model/task/topic/effort mix, hourly heatmap, personal tips, CSV export. IT LEARNS ON YOUR DEVICE • Personal calibration of your plan size and typical answer lengths, from your own usage. • A small on-device task model that learns from your corrections. PRIVACY Everything is computed in your browser. Your messages, attachments and chats never leave your device. No server, no analytics, no tracking. It runs only on claude.ai with your own session and never sends messages or changes settings for you. It asks for your consent before reading anything. LANGUAGES Interface in 10 languages: English, Turkish, German, French, Spanish, Italian, Portuguese, Russian, Japanese, Chinese. Message analysis (task, topic, model and effort suggestions) currently works for English and Turkish messages; in other languages you get the token and limit forecast without suggestions. Open source (GNU AGPL-3.0): https://github.com/H-ENES1/ai-assistant-claude Website: https://h-enes1.github.io/ai-assistant-claude/ Token Forecast is an independent project and is not affiliated with Anthropic. "Claude" is a trademark of Anthropic PBC. Anthropic does not publish exact limit sizes, so forecasts are estimates that improve as the extension measures your usage.