main.cpp runs on a Teensy. It reads color data over USB and forwards it to one or more LPD8806-based LED strips. The format is [r byte, g byte, b byte]*32. Each color can have a max value of 127. Compile with make.
led_driver.py runs on a regular computer (tested on Linux and OS X). It reads audio from a microphone, does FFTs on it, makes the FFT data look pretty, generates cool colors, combines the FFT and color data, and sends it to the Teensy.
More info at http://yager.io/LEDStrip/LED.html
You can technically use this software with any microphone interface and it will work. However, to get clean, perfect-quality digital audio, you will have to use a nonstandard setup.
On OS X, you can use Soundflower or something like that.
I used Ubuntu Studio. It uses JACK, which provides a nice monitor interface that perfectly replicates the digital signal sent to the speakers. To use some ALSA software (mpd) with Ubuntu Studio, I used snd-aloop. I'm pretty sure I read this article. It's a bit of black magic. There may have been other steps to make this work.
If your Linux distro uses PulseAudio, apparently it's pretty easy to set up a monitor interface. Check out #1