modelmux

You'd reach for this when you want Claude to automatically delegate subtasks to different LLMs based on what each does best. Instead of manually switching between models, modelmux routes requests to Codex for code generation, Gemini for multimodal tasks, and Claude for reasoning, all from a single interface. It assumes you've got the respective CLIs installed and configured locally. The smart routing logic decides which model gets which job, so you can treat multiple LLMs as a unified resource pool. Useful if you're already juggling API access to several providers and want to stop context switching between them.