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...)

a mod for Claude Code that uses Haiku 5.5 or Jev to figure out the best model and effort for the task. (it is smart and uses earlier messages to assess if a switch is good or if it’s better to stay on the current one. what it does and how: picks Haiku, Sonnet or Opus and the effort for each prompt warns when the prompt cache goes cold warns when the chat gets swamped Effort is an estimate, but I’ve found it picking the effort I would’ve done almost every time why it works: it reads and understands your current work and task, so it doesn’t flip models mid-task Changing effort on Opus and Sonnet doesn’t rebuild the cache on Fable it does, so automatic switching is off there. Changing effort would cost more than it saves it also offers other features such as smart handoff (a recommendation if it notices a change in the task), this also lets you use your current skill as the hand off template if you have any. I built it with Claude Code. Claude wrote most of the code and I tested it, chose what it should do and fixed what felt wrong. Haiku 5.5 does the judging at runtime. It’s free and open source (MIT). Built a little website for it where you can install, setup is three steps: heycubit.github.io/effortless If anyone has any questions and concerns please let me know! I’m happy to talk about it :) I built it because I kept forgetting to switch models, so I ran everything on high setting. The “judge” is Haiku, with Jev as an option, and it reads your recent messages before it changes anything.