Emphasis on knowledge and information is one of the challenges of the 21st century to differentiate the intelligent business enterprises. Enterprises have to develop their organization in order to capture, manage, and use information in a context of continually changing technology. Indeed, knowledge and information are completely distributed in the information network of the company. In addition, knowledge is, by nature, heterogeneous, since it is provided from different information sources like the software, the technical report, the meeting statements, etc. The authors present in this article the architecture of a multi-agent system, which allows the capitalization of the distributed and heterogeneous knowledge. They then present how the agents help business experts to design ontologies in detailing this problem and how the agents extract knowledge from different user databases by using a semantic approach.