In this work a software agent based on immune mixed selection is developed. The software agent works in a two dimensional environment represented as a grid. Information about the normal configuration of a path in the environment is considered as the training set of the software agent. The goal of the agent is to learn information about the environment in order to be able to detect any change once it has been trained. A set of detectors, which will characterize the information of the environment, is generated through a learning process based on immunology. Some of the detectors will characterize the positive space (self) while the remaining ones will characterize the negative space (non-self). Some experimental results are presented and compared to other two immune approaches, one based on negative selection and the other on positive selection.