caveman-compress

To reduce token usage in markdown memory files, apply deterministic compression to natural language content while preserving core substance. The tool validates output and saves the uncompressed original as a backup file.

You can compress CLAUDE.md, todos.md, and memory notes across projects to minimize token costs during sessions. The process applies deterministic compression to natural language and creates an explicit backup.

Deterministic natural language compression for markdown CLI workflows shrinks token load by algorithmically condensing text without altering meaning. It detects natural language, compresses it, and validates the output.

To run compression on a markdown file from the command line, execute the compression script with the target filepath. The process compresses the natural language content and automatically generates a backup.

Markdown token compression does not destroy original project memory notes. It keeps a human-readable backup by saving the original file as FILE.original.md before applying the compressed changes.

A limitation of compressing markdown memory files for token reduction is that it targets natural language specifically. Output is validated with targeted fixes if needed to ensure content remains intact.