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.

The PH Community (ph) is a forum and resource hub for PHP developers, providing discussions, tutorials, and insights on PHP programming and web development. Developers can ask questions, share knowledge, and collaborate with fellow PHP enthusiasts on topics such as PHP frameworks, CMS platforms, and database integration. Additionally, PH Community covers PHP events, job opportunities, and community initiatives, fostering a supportive and engaged PHP developer community.

Product Hunt

Caveman is a free, MIT-licensed plugin for Claude Code and other AI coding tools that reduces output token usage by ~75% by training the model to skip verbose filler phrases and respond concisely. It supports four intensity levels (Lite, Full, Ultra, Classical Chinese), includes terse commit message and one-line PR review modes, and a compression tool that rewrites CLAUDE.md files to cut ~46% of input tokens. Compatible with Claude Code, Cursor, Windsurf, Copilot, Cline, and 40+ other tools. Install is one line. The project has 24.9K GitHub stars.

Why use so many tokens when a few do the trick?

We needed a GitHub repo for this? Weird. I’ve literally had these Custom Instructions since GPT4, does exactly the same thing. And much, much cooler, IMO.
“Respond purely like the Ship’s Computer from Star Trek. Minimal to no conversation or niceties, blunt and purely factual, no compliments. Be critically helpful, and matter-of-fact.”