This study aimed to develop a K-12 classroom observation protocol to assess K-12 teachers' implementation of science, technology, engineering, and mathematics (STEM) integration. The intended purpose of the observation protocol is for researchers to examine how K-12 teachers implement the STEM integrated curriculum. Based on research on STEM integration, the protocol covers eight categories: "Define," "Learn," "Plan," "Try," "Test," "Decide," "Improve," and "Instructional Strategies." The protocol was reviewed by a panel of integrated STEM education experts, engineering design experts, and educational assessment experts (n = 15) and revised accordingly. To test the reliability of using the protocol, two raters applied the protocol to videos of 12 lower-level elementary teachers located at seven schools who implemented an integrated STEM curriculum over a total of 196 class periods. The inter-rater reliability Gwet AC1 coefficient was 0.73. The inter-rater reliability of each category's coding was also calculated and reported. We discuss the implications and limitations of the protocol.