caveman-compress

You can reduce token usage in markdown memory files by compressing natural language text into caveman format. This process shrinks file size while preserving URLs, code blocks, headings, and file paths.

Caveman compression for markdown tokens is a technique that converts natural language memory files into a compact caveman format. It safely backs up original files while enforcing a 500KB size limit and refusing sensitive paths.

Yes, you can compress CLAUDE.md size without losing code blocks and URLs by using caveman compression. It specifically preserves file paths, headings, code blocks, and URLs while compacting the surrounding natural language text.

Markdown compression refuses to process sensitive filenames or paths to ensure safety. It enforces a 500KB limit, validates output with retries, and creates a backup of the original file before compressing.

To safely compact long project notes for memory storage, run the caveman-compress CLI on your markdown files. It creates a backup as <filename>.original.md and validates the compressed output with retries.