TokenPropulsion is an open-source Python toolkit for safe, auditable LLM prompt optimization. It combines structural and lexical compression, validation and reproducible benchmarks. Early synthetic tests show 29.56% fewer tokens using cl100k_base. Research alpha.