Architects use graphic conventions in their drawings that have meaningful content to the design task. In previous work, such well-defined sets of graphic entities have been identified and defined. These sets are called graphic units. In this paper, we discuss how graphic unit recognition in drawings can take place using a multi-agent systems approach. This approach seems promising as singular agents may specialize in graphic unit-recognition, and multi-agent systems can address problems of ambiguity through negotiation mechanisms. We present an agent framework for this purpose, how it connects to the theory of graphic units, and how agents for recognizing graphic units are defined. The paper ends with a discussion of current findings and future work.