In this work a proposal of a recommendation system based on multiagent is made, which shows the architecture designed, the server used for the development of the multiagent system, as well as the communication between the necessary agents to carry out a recommended route. The implemented proposal allows to carry out the recommended type of route or bookmarks to users about historical and cultural memories. By means of neural networks it is determined if the proposed route for the user is suitable, according to their tastes or preferences or if it is necessary to perform a feedback of the information to improve the following recommended routes. To generate the recommendations, we used the data set of a parallel project, hosted in Amazon Web Service (AWS). An online user database API was also used, which allowed the storage of the information. For the visualization the project was connected to a prototype of an augmented reality system, where the functionality of the system and the obtained results are seen.