Exploring 3D urban datasets is often slow and complex due to occlusion and the need for manual viewpoint adjustments. We introduce a neural field–based, view-driven approach that encodes environments into an efficient implicit representation, enabling both direct queries (like visibility or solar analysis) and inverse queries (to find suggested views). Validated through real-world case studies, our method supports urban analysis tasks such as facade visibility, outdoor space evaluation, and assessing new developments.