paleo: open-source toolkit to cut token costs for AI agents
paleo is an toolkit designed to reduce the amount of tokens (the units AI models use to process text, which determine cost) that LLM-based agents consume. It compresses an agent's output, trims unnecessary parts of the , and lets users set a hard cap on how many tokens can be used.
It's built to work across multiple setups, including Claude Code, Codex, Gemini, and Hermes. The project is published as an by a developer or team called mocasus.
Key points
- Compresses agent output, trims , and caps
- Works across multiple agent setups: Claude Code, Codex, Gemini, and Hermes
- Published as an project on GitHub by mocasus