We use multimodal conversation analysis to investigate programming activities as embodied and interactional phenomena. Drawing on the theory of embodied cognition, we analyse video-recorded interactions of Finnish pupils aged 9–10 years who are programming block-based codes for educational robots for a collaborative story animation project. The analysis focuses on how the pupils negotiate directional instructions, transform them into code, and implement their programming. The analysis reveals i) how the coding platform and its features constitute a situated problem and material space that the participants structure through embodied and social negotiation, ii) how the participants use their bodies as resources for cognitive reasoning and communicating the contingencies of the activity and iii) engage in object manipulation for testing and accommodating what is possible with the space and materials. The findings also indicate that a trial-and-error approach in which pupils test, observe and refine their code is essential to activity engagement and for developing programming and computational thinking skills.