Point-of-interest (POI) recommendation is a paramount task that spans two important genres. One is the personalized POI recommendation that integrates location, time, comments, social circle, etc. The second dimension is group-based POI recommendation, which includes suggesting POIs to groups of people. Recommending to the group is quite a challenging task as it requires satisfying the needs of all the group members and catering to the social dynamic. In this paper, we propose a multi-agent system (MAS)-based group POI recommendation (MAS-POIRec) framework where each agent represents the users taken in groups and negotiates iteratively based on the additional information of each user to reach a consensus. The MAS approach can harness the LBSN data to provide more suitable recommendations that are in the users’ best interests. In addition, to limit the amount of information circulated between the group agents, we adopted a beta reputation management system modified to include social connections and temporal weights. The final beta reputation scores are combined with the confidence level score of each agent to establish the reliability of each agent with the other agent. This reliability computation is then used to recommend POIs for the entire group. The experiment on a real-world dataset, Brightkite, demonstrated substantial satisfaction among the agents and fairness in group-based recommendations.