Basically, it makes a graph of people you've friended who have friended each other. The size of the dots is scaled to the number of mutual friends that person has. So, people with a BUNCH of friends will have a bigger circle than people with fewer.
Here is a small example from a cluster on my friendslist:
The functions in the package also provide some nice functionality for on the fly processing (ie, figuring out who do X and Y have friended in common). There's a fdata download and caching manager, and it rate limits its requests, so you don't have to be concerned about being hard on the servers when running it.
It can be found on BitBucket. I'm not too familiar with Python yet, though I have been enjoying it immensely, so anybody who Knows Better on how to do things is free to give me a guiding poke.