Proteins are the functional and evolutionary units of cells. On their surface, proteins are sculpted into numerous concavities and bulges, offering unique microenvironments for ligand binding or catalysis. The dynamics, size, and chemical features of these cavities are essential for the mechanistic understanding of protein function. Here we present CaviDB (https://www.cavidb.org), a novel database of cavities and their features in known protein structures, which integrates the results from commonly used software for cavities detection with protein features obtained from sequence, structure, and function analyses. Additionally, each protein in CaviDB is associated with its corresponding conformers which help to analyze conformational changes in cavities as well. We were able to characterize a total number of 16,533,339 cavities, 62,0431 of them predicted as druggable targets. CaviDB contains 276,432 different proteins, with information about all their conformers. It also offers the capability to compare cavities and their features from different conformational states of the protein. Furthermore, we have recently added the available models from the AlphaFold database versions 2 and 3, which allow further cavity explorations and comparisons. Each entry information is organized in sections, highlighting the general cavities descriptors, including the inter-cavities contacts, activated residues per cavity, the information about druggable cavities, and the global protein descriptors. The data retrieved by the user can be downloaded in a format that is easy to parse and integrate with custom pipelines for protein analysis. CaviDB aims to offer a comprehensive database for use not only in different aspects of drug design and discovery but also to better understand the basis of the protein structure-function relationship better. With its unique approach, CaviDB provides an essential resource for the wide community of bioinformaticians in particular and biologists in general.