The paper presents in brief a methodology for development of tools for knowledge-based search in repositories of digitized manuscripts. It is designated to assist the search activities in collections that may enlist XML documents which should be catalogue descriptions or marked-up full texts of mediaeval manuscripts. The suggested methodology is directed to the development of software environments that will be able to deal with relatively complex user queries containing words or phrases that are considered as domain concepts. The emphasis in this methodology falls on two main types of activities: development of proper ontologies describing the conceptual knowledge relevant to the chosen domain(s) and development of proper intelligent agents for search and processing purposes that are able to retrieve and filter documents by their semantic properties. Some considerations related to the implementation of our methodology are presented as well. The first version of a software tool for knowledge-based search in repositories of digitized manuscripts is discussed. Some results of the application of the tool to a collection of approximately 800 descriptions of mediaeval Bulgarian manuscripts stored in Bulgaria are analyzed.