Caveman: make Claude talk like a caveman to cut output tokens
TL;DR
Caveman is a community skill/plugin for Claude Code that makes Claude reply in terse, telegraphic style. Generated code is unchanged; only the prose is cut. The project cites about a 65% cut in output tokens, but total session cost drops much less because output is only part of it.
Modes
| Mode | Description |
|---|---|
| lite | Removes filler but keeps full grammar |
| full (default) | Fragmented sentences, dropped articles |
| ultra | Maximum compression, note-style |
Variants also exist for short commit messages, one-line PR reviews, and compressing CLAUDE.md to save input tokens.
Install
claude plugin marketplace add JuliusBrussee/caveman
claude plugin install caveman@caveman
# then type /caveman in a sessionWhen to use it — and when not
| Use it for | Avoid it for |
|---|---|
| Repetitive work you already understand (small fixes, familiar refactors) | Learning a new framework, when you need explanations |
| Long sessions with many turns | Architecture decisions, debugging unfamiliar errors |
| Commit messages, internal notes | Content for colleagues or clients |
Further reading: Caveman mode analysis