This paper proposes a framework for animating the behavior of virtual agents that can reproduce several human behavior features with a minimal input from the animator. While this work is not actually concerned about obtaining a high-quality animation in terms of motion, illumination, textures or any other rendering features, we focus on modeling and simulating the human behavior so that the virtual agents behave in a realistic way from the point of view of a human observer. The framework is illustrated by a simple but still interesting example: a virtual park where the virtual agents can walk, chat, play, rest, do physical exercices, interact with others, etc. The system is flexible enough so that any new goal, feeling, sense, internal state, knowledge or social habit can easily be added without modifying the underlying structure. Some examples of human behavior features such as the competition, anxiety, etc. are also described.