This chapter presents an experimental development of an on-demand system for human settlement mapping using the Landsat archive. The system was implemented with automated algorithms of satellite data selection for user’s preference and human settlement mapping using a machine-learning-based method, called Learning with Local and Global Consistency, with a modification for satellite data application. In addition, the system was developed with a web-based user interface for user’s parameter inputs to customized data products delivered by a geospatial data standard of Web Processing Service (WPS). The developed system processed 838 scenes of Landsat data for 1990, 2000, 2005, and 2010. The results visually showed urban growth with settlement development in a fair quality although there were considerable overestimations for 1990, indicating that time-series consistency of settlement development would improve accuracy of the output. The on-demand system provided an interface of receiving user’s parameters embedded in URLs to WPS. As the WPS is a standard compatible with other mapping standards, the on-demand system will be integrated with user-friendly WebGIS interfaces to appeal more to end users who are key players in achieving Sustainable Development Goals.