This paper illustrates the design and implementation of FAIN (Filtering Agent for Internet) based on the Gaia methodology and the AUML notation. The proposed system generates and maintains a user profile by learning from examples that are judged and classified by the user during the training stage. A learning algorithm which is based on the Baldwin effect that combines learning with evolution is applied to judge an online stream of documents and filter it according to the learned profile. The illustration of the design and implementation of FAIN has shown the applicability of Gaia complemented with the AUML notation to the modeling of agent-based system.