Iterative Clustering procedure for bidimensional data

Hi, I need to cluster bidimensional data represented by start_time-duration of a generic activity. My purpose is to detect only change in data, so I have not a predefined number of cluster but if a new cluster is generated, I need to signal the event (so, Kmeans clustering can't be used for my task)

I'm searching for an iterative clustering technique able to cluster the data in a FOR cycle (when a new data is available...evaluate the distribution of points and count the clusters...)

Any idea?