What you get

Reduce output tokens by an average of 65% while maintaining 100% technical accuracy.

Reduce AI Token Usage by 65%

Added 2026-07-26

Reduce AI output tokens by 65% by adopting a concise, caveman-style communication mode without sacrificing technical accuracy.

VerifiedEngineering
Reduce AI Token Usage by 65% preview

Caveman is a plugin for Claude Code, Cursor, Windsurf, and 30+ other AI coding agents. It strips away filler text and compresses AI responses into essential points, significantly reducing token costs and improving readability. The tool supports multiple compression levels and includes features for optimized commit messages, code reviews, and session token usage statistics.