Traditionally Intelligent Tutoring Systems (ITSs) are built with modular architecture. Such ITSs consist of four traditional modules, namely tutoring module, expert module, student diagnosis module and communication module. Three of these modules correspond to the main types of knowledge (pedagogical knowledge, domain knowledge and knowledge about the learner) used in ITSs, so allowing to build components that are using only one type of knowledge. Still, modular architecture does not provide sufficient modularity for complex ITSs. To facilitate modularity and change management, lately distributed technologies like services and intelligent agents are used to develop ITSs. Previous research has concluded that agent based ITSs mainly use the same approach – they implement traditional modules as sets of agents. Customizable set of agents that can be used in various ITSs has been defined. Implementation of modules as a set of distributed components allows to keep the traditional idea of modules. Different types of knowledge are separated from each other. At the same time, the system can be distributed and highly modular. A few multi-agent architectures for ITS development have been published. These architectures mainly consist of agents from the abovementioned set of agents. The architectures are closed in the sense that system’s functionality can not be changed just by adding/removing components from the system. As a consequence multi-agent architectures for ITS development disable one of the advantages of distributed technologies – possibility to change system’s functionality by just adding and/or removing distributed components from the system. As of author’s knowledge no specific service oriented architectures (SOA) for ITS development exist, except the ones presented further in the chapter. So there is a lack of open ITS architectures, despite ITS being a system that may need frequent changes due to the changes in the particular course or adaptation to new courses. To realize intelligent tutoring various types of learning materials and problems have to be presented to the learner, moreover it should be done intelligently enough. For example, the system has to be capable to analyse each student’s action during the problem solving. Thus each new type of problems may need corresponding code to handle it. Types of problems differ from course to course as well as might change in case the course is changed. So, the system should be open for corresponding new components. Actually, typical ITSs must be open for certain types of components handling new types of problems, materials, feedback, etc. Presence of specific distributed and open architectures would facilitate development of open ITSs. The aim of the chapter is to propose open ITS architectures, using two distributed paradigms – intelligent software agents and services.