In this paper, we present a genetic algorithm-based approach to designing specific self-assembling structures that act as building blocks for assembling more complex objects. Our simulated environment models two-dimensional square blocks as autonomous agents. A specific class of multi-agent self-assembled products is termed "enzymes", which are capable of producing other multi-agent self-assembled products whilst themselves remaining unchanged. The global goal of our self-assembly process is to produce large stable structures by evolving the enzyme and block parameters. Our experiments show that the "enzymes" can interact with the other blocks to produce a wide variety of stable self-assembling structures.