Last released Nov 22, 2026
Cut LLM input tokens: compress tool outputs, logs, JSON and chat history before each call. Runs in your process; your data stays on your machine.