Video cameras are increasingly present in our every-day environment. As a result their monitoring by human users has become basically unrealizable. At the hands of this problem, automatic processing technologies have been gradually developed and now offer reliable analysis methods. Notwithstanding, the integration and use of analysis results are currently barely tackled. Furthermore, the development of a system allowing these kinds of activities calls for particularly adapted technologies. The feasibility demonstration of such systems is consequently a daring challenge. In order to take it up, state-of-the-art technologies are to be combined in an efficient way. Traditional object-oriented and data storage approaches may fail to offer the required flexibility and scalability. Less common and mature technologies were thus explored. The knowledge processing mechanisms we set up are based on agents. Agents are notably autonomous and reactive entities exhibiting very adaptive behaviours. The knowledge structuring and storing resulting from varying video analysis levels are based on Semantic Web technologies. The RDF graph structure appears to perfectly support our needs of a priori undefined metadata schema. Moreover the integration of these components in a distributed and possibly heterogeneous environment raises concerns regarding knowledge exchange and sharing. The system we developed intends to handle these concerns by associating computer science specific concepts in order to enable real-time video analysis sharing.