I built TokenTrim to trim the padding from AI prompts ("Could you please ", "make sure that you ", repeated instructions) before they reach ChatGPT, Claude or Gemini. It cut my own token usage by about 40%.
The rewriter is rule-based and runs locally, so it's fast and private, but rules can be wrong. If you try it, I'd love to know one thing: paste a prompt where it cut something you actually needed, and what mode you were in (Light / Balanced / Max).
Every example here goes straight into the next version's rules.
Free Chrome extension that compresses your prompt before it reaches ChatGPT, Claude or Gemini. One click, see the diff, send. Runs locally, stores nothing.
