Grid tools currently available are most often geared towards running batch-oriented dataset analysis jobs on a set of distributed heterogeneous computers in a production setting where the data is assumed to be in files, and do not automatically collect the results from multiple analysis jobs. This is insufficient for interactive dataset analysis, where complicated communications between the analysis jobs and the scheduler are required for interactivity and where datasets are abstracted from files, collections of files, database objects, and even virtual datasets for which only a generation algorithm exists. We describe a prototype for a system based on a set of grid services to provide interactive dataset analysis that extends the Globus Toolkit grid tools. An existing Java-based dataset analysis and visualization client tool are used to access the grid service and provide the interactive user interface for the client.