Token Pulse

Description

A live counter that estimates the tokens of your prompt or conversation, the input cost, and how much of the chosen model's context window you are using. It reads the input text from the page on demand (activeTab) or works on pasted text. The estimate uses a robust heuristic (~chars/4 with punctuation and whitespace adjustments): it is not the model's real tokenizer but stays within a typical ±10-20% tolerance. Ships with a bundled table of models and context windows, from a free on-prem local model up to 2M tokens. 100% offline, zero telemetry.

Features

• Live token count as you type or paste text
• Estimated input cost in USD for the selected model
• Context window usage percentage with progress bar and status (ok/warning/overflow)
• Remaining free tokens and over-budget warning
• Reads the input text straight from the page (activeTab, on-demand)
• Bundled table of models and context windows (from on-prem local to 2M tokens)
• Remembers the selected model across sessions (local storage)
• 100% offline, no data leaves the browser

Required permissions

• activeTab
• scripting
• storage

Screenshots

Supported browsers