The development of learning contents is a collaborative process in which, users with different backgrounds, experiences and points of view can take part. This is a two-step process: first, a negotiation and evaluation of ideas amongst participants and second, composition of learning contents as result of that negotiation. However the collaborative development support in current learning contents authoring tools is scarce, since they usually provide a uni-personal vision of the creation process. This work presents a multi-agent system for the collaborative authoring of Reusable Learning Objects (RLO), which key features are: (1) collaboration stakeholders are represented by agents; (2) agents message exchange is coordinated by means of a negotiation protocol; (3) agents rationality for proposal evaluation is modelled in terms of preference relationships; (4) a group of assistant agents trace all negotiation events, so allowing the whole creation process to be audited, whilst keeping a repository to control changes accomplished into the RLO corpus, thus allowing automated consolidation of negotiated results.