Assembler is a computational tool that utilises automated combinatorial growth for the design and study of architectural assemblages. It focuses on generating heterogeneous spaces through the iterative addition of duplicated parts from a limited set. This method leverages combinatorial design to achieve complexity through the arrangement of repeatable basic parts. Unlike existing tools that rely on random choice or predefined patterns for rule application, Assembler allows for controlled and adaptable design by employing deterministic, rule-based criteria where parts function as computing elements, mobilizing degrees of distributed, autonomous computing. Assembler’s conceptual framework draws from DeLanda’s formulation of assemblage, emphasizing the importance of both parts and their relationships in generating emergent qualities. Parts and connections are treated as computing objects representing the geometry of reserved space and its connectivity. This allows for data (in both numerical and geometrical form) to be embedded within the parts and the overall assemblage, enabling computation and automated decision at local, regional, and global scales. The user has control over various aspects of the design process: the design of parts and connectivity, the system’s heuristics (presets or customized via programming), and environmental information, encoded as a discrete tensor field to embed multidimensional data. This results in a data-rich, computable structure of voids, where geometrical, topological, and numerical data are readily accessible for integration with other analytical and computational tools. Additionally, Assembler’s granular computation and decision-making processes are programmable with bespoke methods, including potential future implementation of Deep Reinforcement Learning strategies for goal-based design.