The subject of this paper is a pilot project devoted to creation and use of a retrieval agent in educational activity of universe, that provides to students, teachers and professors a means of contextual search of educational resources on the Internet. It describes an agent that provides a process for the extraction of training materials and other data from the World Wide Web and their systematization. The primary search of content on the Internet is carried out by the known search engines (Google Ajax Search, Yandex, Yahoo, Mail.ru). An adaptive technology of sorting and clustering of the found network content on the basis of the Pareto dominance relation under the joint consideration of many aspects (relevance, pertinence, etc.) has been offered and implemented. Fuzzy text comparison is performed with use of the Levenshtein distance. Key architectural, technological and design solutions are presented. Samples of the user interface are shown. Results of testing the software productivity ware are given. The article is aimed at a wide range of professionals who are interested in the application of contextual search on the Internet, as well as university students and teachers, who specialize in the field of informatics and computer science.