Abstract Efficient resource management in dynamic cloud computing is receiving more and more attentions. Many works are still ongoing, since federated cloud computing is an open environment. We present in this paper a flexible model, new cloud operators can join the federation or leave it. The model integrates interactions between broker agents organization, permitting a multi-criterion migration of a submitted customer request. To implement the presented flexible model, we propose AMACE (agent based multi-criterions adaptation in cloud environment) a run-time self-adaptive approach, oriented agent in dynamic cloud federation environment. The proposed approach is multi-criterion, where a various and no limited number of parameters can be considered during the self-adaptation strategy (i.e. computing the load balance of the mediator agent and geographical distance “network delay” between the customer and provider…). In addition, AMACE cares about maintaining run-time coherence in optimization evolution in order to guarantee a coherent system at all moment during system execution.