In many countries, agriculture is a means of development. Several structures are set up as cooperative, economic interest group (EIG), etc. However, these structures have very little interdependency that can allow them to go beyond their limits such as access to finance, access to national and international markets, etc.. This raises new challenges of collaboration between structures that we can consider each structure as a coalition. This work provides a coalition’s migration (dynamic merging and splitting coalition) mechanism that allows auto-stable coalitions of self-interested agents in uncertainty context. Specifically, we assume uncertainties and interdependencies on tasks, agents and resources. We address the case in which agent and environment inherent uncertainties prohibit the computation of coalition stability ahead agent’s goals. Facing such context, we propose a core-stable and auto-stabilizing anytime coalition formation mechanism which we denote as DMS (Dynamic Merging and Splitting). The mechanism arrives at a stability, maximizes social welfare, and converges gradually to near optimal results. DMS combines game theory methods and the laws of probability. Our experiments and their analysis demonstrate the efficiency of DMS.