Why use so many token when few do trick?
Caveman cuts ~75% of Claude's output tokens without losing technical accuracy. One-line install for Claude Code, Cursor, Windsurf, Copilot, and more. Four grunt levels, terse commits, one-line PR reviews, and input compression built in. 24.9K stars.
Caveman provides /caveman-review (one-line PR comments with specific line numbers and fixes) and /caveman-commit (conventional commits ≤50 char subjects). Developers get focused, scannable feedback instead of verbose explanations, improving code review velocity.
Caveman includes /caveman-compress skill to rewrite memory files (CLAUDE.md, project notes) into compressed caveman-speak, cutting ~46% of input tokens. This allows agents to retain more project context across sessions without hitting token limits, enabling longer-running development workflows.
Caveman compresses AI agent output by ~65-75% by enforcing terse, caveman-style communication while preserving technical accuracy. Users install the skill/plugin once and trigger with /caveman command, immediately cutting output tokens and associated API costs without sacrificing code quality or correctness.
By reducing output verbosity, Caveman enables ~3x faster response times from AI agents. Shorter responses mean quicker feedback loops, allowing developers to iterate faster on code generation, debugging, and review tasks within their existing IDE or terminal.
Caveman's /caveman-stats command displays real session token usage, lifetime savings, and USD cost reduction. Output includes a tweetable summary and updates a statusline badge in Claude Code, giving developers immediate visibility into cost benefits and ROI of the compression skill.