This paper presents a novel type of intelligent agent with a multilingual natural language interface, which retrieves information from within a Web site. This agent, named JaBot after the fact that it is a bot which has been programmed in Java, has been designed and developed by the authors in an attempt to solve common Web site problems related to information retrieval. JaBot runs quickly and efficiently, and rather than running directly on the Web site pages, it is connected to a lexical semantic map. This map is based upon the contents of the Web site in question together with other associated linguistic knowledge.