MemTensor / MemOS

Self-evolving memory OS for LLM & AI Agents: ultra-persistent memory, hybrid-retrieval, and cross-task skill reuse, with 35.24% token savings and DeepSeek Harness support.

zdk / lowfat

lowfat - slim your command output. strips noise, saves tokens.

jfrog / boost

Save tokens. Maximize context, Safely

Fabian-Galvez / DensePack

Packs text files into small images that an AI model reads for about half the input tokens. Includes a Claude Code plugin, a right-click tool and an HTML app.