With the expanse of Internet, web programmers have wide choice of web services available to them. A need arises for automatic discovery of required web services and construction of an appropriate sequence of invocation thereof. In this paper, we present a framework for automation of this task based on currently emerging technologies such as ontological knowledge bases, OWL, OWL-S, WSDL, Description Logic (DL), etc. Background knowledge ontologies are created based on which semantic meanings of web services can be given through OWL-S. An agent employs OWL-S API to extract web service metadata, and applies a DL inference engine, called Racer, for reasoning with the metadata with respect to given background knowledge. Reasoning tasks performed by Racer include profile matchmaking, input/output subsumption testing, and preconditions/effects analysis, which are basic mechanisms for web services discovery and invocation planning. A prototype system has been implemented. © 2006 IEEE.