The growing specialisation and ever-increasing inter-relationships in medicine today leads to multidisciplinary and distributed decision-making. Supporting these decisions raises some challenges to current software frameworks and knowledge models. In this paper, a Multi-Agent Framework and its accompanied knowledge model is put forward. The knowledge model originates from published clinical guidelines and drives agents for decision support in an adaptive manner. The primary components of the model include agent goal structures representing state-transition along care pathway, agent interactions composed by recurrent patterns, and agent executable rules for individual planning and argumentation. A number of framework services facilitate the enactment of the knowledge model by agents, and the approach is illustrated using a case study of Triple Assessment in breast cancer.