Background: Coding has been added to school curricula in several countries, being one of the necessary competencies of the 21st century. Although it has also been suggested to foster the development of several cognitive skills such as computational thinking and problem-solving, studies on the effects of coding are very limited, provide mixed results, and lack causal evidence. Aim: This study aims to evaluate the impact of a learn-to-code programme on three cognitive skills in children: computational thinking, fluid intelligence, and spatial orientation, using a randomized trial. Sample: One hundred seventy-four (n = 81 girls) 4th-grade children participated in the study. Methods: Children were randomly assigned to one of the three 10-week learning programmes: learn-to-code (treatment of interest), mathematics (another STEM-related comparison treatment), and reading (control). Children responded to paper-pencil computational thinking, and spatial orientation measurements, and face-to-face matrix reasoning task at pre- and post-tests. Results: Results showed that children's computational thinking scores increased significantly only in the learn-to-code condition. Fluid intelligence significantly increased in all conditions, possibly due to a practice effect. The spatial orientation did not improve in any of the conditions. Conclusion: These findings suggested that learning to code can be selectively beneficial for the development of computational thinking skills while not effective for spatial reasoning and fluid intelligence.