hogashi/agent-b2b ? reverse-engineered prompt

Build me a small local web app that plays a beat and changes what I hear based on whether my AI assistant is working or waiting for me.

I want a simple page with a power button, a sequencer, track controls, BPM and volume sliders, and a way to save and load my settings as JSON. The main idea is that when the agent is thinking, one fuller music layer plays, and when it is my turn, a simpler beat plays. It should react to local status changes from a tiny server on my machine, with everything staying on localhost and nothing sent online.

Please make it easy to run from the terminal, open the browser automatically, and hook into Claude Code with simple local commands so the app knows when to switch between the two states. If you need to check current docs for any browser or audio details, feel free to look them up online.