We extend our previously developed agent-based (AB) algorithm to the study of the self-assembly of a fully atomistic model of experimental interest. We study the 2D self-assembly of a rigid organic molecule (1,4-benzene-dicarboxylic acid or TPA), comparing the AB results with Monte Carlo (MC) and MC simulated annealing, a technique traditionally used to solve the global minimization problem. The AB algorithm gives a lower energy configuration in the same simulation time than both of the MC simulation techniques. We also show how the AB algorithm can be used as a part of the protocol to calculate the phase diagram with less computational effort than standard techniques.