During the last decades many Intelligent Tutoring Systems (ITS) are developed to add adaptivity and intelligence to the e-learning systems. Intelligent agents and multi-agent systems are widely used to implement intelligent mechanisms for ITSs due to their characteristics. The paper presents an agent based ITS for the course “Fundamentals of Artificial Intelligence” named MIPITS. The MIPITS system is based on the holonic multi-agent architecture for ITS development. The system offers learning materials, provides practical problems and gives feedback to the learner about his/her solution evaluating his/her knowledge. The goal of the system is to realize individualized practical problem solving, which is not possible in the classroom due to the large number of students in the course. Thus, the main focus of the system is on problem solving. The system offers three types of problems: tests, state space search problems and twoplayer games algorithm problems. The MIPITS system is open: new types of problems can be added just by including appropriate agents in the system. The problems are adapted to the learner’s knowledge level and preferences about difficulty, size and practicality of problems.