In urban environments, roadside vegetation provides important ecosystem services. Reliable and up-to-date information on urban vegetation is therefore needed as a basis for sustainable urban design and regular tasks such as vegetation maintenance. Mobile laser scanning (MLS), i. e., the use of vehicle-mounted laser scanners, offers strong potential for capturing 3D point clouds of road environments on a large scale at a low cost. In this paper, the potential and challenges of using MLS for vegetation mapping are discussed. To lay a foundation for MLS-based inventories of roadside vegetation, a concept for the automatic detection and analysis of vegetation in MLS point clouds using deep learning is presented. The proposed workflow covers vegetation detection and classification, delineation of individual trees, and estimation of tree attributes. In a case study, an initial implementation of the workflow is tested using MLS datasets from two German cities and the results are evaluated through visual inspection. It is demonstrated that the proposed deep-learning approach is able to detect and classify vegetation in MLS point clouds of complex urban road scenes. When delineating individual trees, accurate results are obtained for solitary trees and trees with little canopy overlap, while the delineation of trees with strongly overlapping canopies needs further improvement in some cases. The results indicate that geometric tree attributes such as tree height and trunk diameter can be accurately estimated from MLS point clouds if the accuracy of the preceding processing steps is sufficiently high.