This paper reports on the computational modelling research investigating spatial organisations often associated with unplanned settlements. Such spatial agglomerates are composed of several co-located but autonomous units (agents) that share common facilities and infrastructure (e.g. circulation). Depending on the context, units in the agglomerate represent individual dwellings, apartments or abstract spatial geometry. The paper presents early prototype models that can be interpreted at various scale, and a computational model for generating organic settlement layouts. The originality of the research resides within a new multi-agent algorithm for creating spatial organisations. The agglomeration process benefits from two distinct generative design strategies – self-organisation and adaptive development strategy. While the self-organisation accounts for the emergence of the global structure in the agglomeration, the adaptive development strategy ensures that the basic environmental and spatial requirements of each individual unit are satisfied.