Urban street networks contain repetitive structures that reflect human needs as cities expand and evolve. To identify and understand these building blocks of cities, we propose the use of graphlet-based methods-that is, focusing on small, connected subgraphs of these networks. Looking at graphlets of up to 4 nodes in the street networks of New York City, we identify local structures such as gridded patches through spatial auto-correlation statistics. This methodology can be quickly applied to any city in the world, helping researchers classify local street structures and identify common urban development trends across many cities.