for hclust the tighter cluster is always the unit, that is then end leaf is always made of one observation. It will split the tree until this level using one method, in R you have 5 or 6.
Now if you want some clusters you have to cut the tree at a certain distance, in your case if you cut at height 3 you have 2 clusters or if you work with R you tell how many clusters you want using cutree() which will give one array the same size as the date set with the cluster reference.
If you cut with height then you have to transform you hierarchical representation (result of hclust()) into a dendrogram and then use cut().
If you want some automatic clustering with other distances and methods (single linkage, double linkage) and you work in R then check NbClust(). This function could give new insight and you can change the method of clustering, only drawback it is slower than hclust().
Hope this help.