The aim of this study was to develop a coding scheme rooted in the Knowledge Building model, named Conversational Functions for Knowledge Building (CF4KB), to analyze students' interactions in an online undergraduate course. In order to develop the coding scheme, we analyzed students discourse and identified the kinds of "Conversational Functions" students performed. Connecting these identified functions with Knowledge Building Model principles, we developed the coding scheme to analyze knowledge building [In the present article, we used "Knowledge Building" to indicate the theoretical model and "knowledge building" to refer to the activity of the community members.] activity. Applying the resulting coding scheme, two independent judges coded 186 messages posted into Knowledge Forum by 26 students (5 men and 21 women) attending the online course of Educational Psychology at University of Valle d'Aosta to develop the coding scheme. The inter-coder agreements obtained were 77.3 and 78.0%, with Cohen's K = 0.68 and 0.66, respectively for Specific and Global Conversational Functions level. The theoretical implications and possible uses in different contexts for future inquiry of this instrument are discussed.