Many NLP systems are based on lexical data. The development costs of such data are a major drawback in such NLP systems. In order to cut these costs, we adopt a strategy inspired from "open- source" projects to allow volunteers to collaborate in the creation of a multilingual lexical database. For this, we had to specify and develop tools to manage a lexical database containing information complete and detailed enough to be usable for a wide range of applications. This paper presents our project and details the tools, frameworks and structures used to manage such a database. We will also show some research problems still to be addressed in this context.