Caveman Code Caveman Code
Caveman Code is a terminal coding agent designed for token efficiency through a multi-layer compression engine. It functions as a provider-agnostic coding harness that integrates with over 20 LLM providers via OAuth or API keys.
As of August 2026, the project status is frozen. Active development has transitioned to the 'caveman' repository, which focuses on wrapping existing agents to improve cost-efficiency.
As of
- 3-layer token compression (Caveman Mode)
- Autonomous goal loop and read-only plan mode
- Persistent memory via cavemem integration
- Support for 20+ model providers and OAuth
- Claude Code-compatible hooks and configuration
Developers looking to significantly reduce token consumption and costs while using AI coding agents in the terminal.
Users requiring ongoing active development, as the project is officially frozen; users should look to the 'caveman' project for continued support.
MIT licensed, free to use.
An innovative, highly efficient open-source coding agent that successfully pioneered multi-layer token compression; while the specific 'caveman-code' package is no longer receiving updates, its core concepts have been integrated into the successor 'caveman' project.
Is Caveman Code free?
Yes - Caveman Code is Open Source. MIT licensed, free to use.
What is Caveman Code best for?
Developers looking to significantly reduce token consumption and costs while using AI coding agents in the terminal.
Who makes Caveman Code?
Caveman Code is developed by Julius Brussee. It is listed in the Terminal & CLI Agents category on ai.dosa.dev.
Favorite this tool to revisit it later, or Zap it to contribute to the public vote count.
Content on this page is AI-generated. Please verify details with the vendor's website for accuracy.