In this paper we present a generalised version of the classical cluster coefficient, which also can be applied to networks with directed links. This generalisation takes into account more than the immediate nearest neighbours, giving more detailed information about the network structure than the classical version. The introduced concept is compared to earlier generalisation attempts, and it is applied to a directed version of the protein interaction network of the yeast cell S. cerevisiae and networks generated by the growing preferential attachment model of Barabási and Albert. Finally, we give some ideas on how our concept is related to modularity and community structures.