A Reddit post describing a Claude Code mod that uses a runtime judge model to automatically select between Haiku, Sonnet and Opus and estimate reasoning effort per prompt, plus warnings for cold prompt caches and overloaded chats, and a smart handoff feature. The mod claims to read recent messages before switching to avoid mid-task model flips. It is described as free and open source (MIT) with a website for installation, but no repository link or verifiable source is included.
I made a claude code mod that uses Haiku 5.5 (or Jev) to pick the best model and effort for your task. (and more...)
Stars 68Permission footprint L3