To use loader.js, include the following <script> tag in an HTML document, changing the attribute values as appropriate: <script src="/uri/of/loader.js" require="/uri/of/lib#main/module/id"> </script> Since loader.js is a static file, you can load it from anywhere. The more people who load it from the same place, the more likely it is to be cached when a user visits your site. That's always nice. The "require" attribute provides two pieces of information, delimited by the '#' character. The first should be a URI (absolute or relative, whatever you want) pointing to the directory that contains curator.py (or a directory containing all possible paths/to/modules.min.js that you might require). This directory does not have to be hosted at the same domain as loader.js. The second half of the attribute is the absolute ID of the first module to load. The curator.py script is intended to serve as a dynamic handler for 404 errors. When loader.js requests a file called some/module/id.min.js, then, assuming this file does not exist, the web server should delegate to curator.py, which will look for a file called some/module/id.js and wrap its contents in some boilerplate that loader.js understands. In production, the output of curator.py should be saved as an actual file called some/module/id.min.js to avoid the overhead of script execution and fully exploit the web server's static file-serving optimization abilities (gzip, etags, cache headers, &c.). See the included .htaccess file for an example of how to configure curator.py as an ErrorDocument under the Apache webserver.