About Token Workbox

Token Workbox helps prompt engineers and developers count LLM tokens and sketch API cost before they send a request. Paste Markdown or prompts, pick a model, and read tokens plus an estimated USD round-trip cost.

OpenAI models use js-tiktoken for exact local counts. Claude, Gemini, and open-weight models use a local approximation (labeled Estimate) because vendors do not ship current browser tokenizers — not billing-accurate.

Principles

• Local-first: your text stays on your device
• Honest accuracy labels (Exact vs Estimate)
• Markdown-friendly workflows for prompt authors
• No vendor token-count API round-trips