We embraced the learning from nature and back to nature paradigm to develop viable agroforestry scenarios through studying species association in 12 wild yellowhorn (Xanthoceras sorbifolium: a Chinese endemic oil woody plants) communities. We identified 18 species combinations for their suitability as agroforestry mixes where positive associations were detected and thus economic benefits are anticipated. In each wild yellowhorn community, we use nonmetric multidimensional scaling ordination to assess community structure and composition, and the climatic variables that most likely influenced existing species distributions. Next, pairwise and multiple species associations were evaluated using several multiple species association indices (e.g., chi(2), Jaccard, Ochiai, Dice). Generally, all species association indices were in agreement and were helpful in identifying several high valued medicinal species that showed positive and significant associations with yellowhorn. Finally, we proposed several agroforestry species mixes suitable for yellowhorn.