Taps into your browser's Web Audio API output and lets Claude analyze what your app is actually rendering in real time. Installs as Express middleware in your dev server, captures audio from the AudioContext graph (not your microphone), then exposes four tools: capture_audio for recording clips, analyze_audio for signal metrics like RMS, spectral centroid, frequency bands, and clipping, describe_audio for plain English summaries, and diff_audio to compare two captures. Works with raw Web Audio, Tone.js, or Three.js audio setups. Requires a CodedSwitch API key for the analysis backend. Useful when you need Claude to hear that your bass is muddy or your timing is drifting without exporting files or recording through a mic.
claude mcp add --transport stdio asume21-webear uvx webear