The volume of information on the Internet is constantly growing. This fact causes that the search of interesting information becomes a time- consuming task. Generally, a user must revise a big number of uninteresting documents and consult several search engines before finding relevant information. A personalized agent, called PersonalSearcher, that assists the user in finding interesting documents in the World Wide Web is presented in this paper. This agent carries out a parallel search in the most popular Web search engines and filters their result, listing to the user a reduced number of documents with high probability of being relevant to him. This filtering is based on a user profile that the agent builds by observing the user behavior on the Web. The agent uses a textual case-based reasoning approach in order to detect specific subjects that the user is interested in and organizes them in a hierarchy that defines the user profile.