Just tweet your link as you would normaly, but use the #zob hashtag anywhere in your tweet. You will find your link in your zobmark account (http://zobmark.com/twitter_username)
To answer your first question:
We crawl the link, and apply a proprietary semantic algo which uses a statistical method - we trained our algo on a preset universe of web pages to get the relevant correlation matrices
About scaling:
We used node.js (server side javascript language), which is very useful for this type of process.