← All projects
ToolIn Development
Defluff
Token minimizer for LLM prompts
Still being built. Some of what is described here is in progress.
WebChrome Extension
About
Cuts LLM token cost by 15–55% before the prompt leaves your machine. A local tokenizer strips filler, dead whitespace and repetition while keeping the meaning, and shows you the diff before you send.
Highlights
- Three modes: light, aggressive, semantic
- Runs entirely in the browser
- Token diff preview before you send
- Works with Claude, ChatGPT, Gemini and Perplexity
Tech
TypeScriptWebAssemblytiktoken
Platforms
WebChrome Extension
