4 tools matching “token-saving”
Context Mode
A free local plugin that sandboxes tool output and gives the agent only the part it asks for — median reply 99.4% smaller.
claude plugin marketplace add mksglu/context-mode
Claude SkillsFreeActive33.7K
Headroom
A local proxy that compresses tool outputs, logs and history before they reach the model, cutting token usage.
CLI ToolsOpen sourceActive79.2K
Caveman
Terse caveman-style answers and smaller context reads that cut token usage — with an honest ~3% median gain.
npx @caveman-ai/cli@latest
Claude SkillsOpen sourceActive118.4K
PonyTail
Makes your agent write 54% less code — like the laziest senior dev in the room.
/plugin marketplace add DietrichGebert/ponytail
Claude SkillsOpen sourceActive187K