The COVID-19 pandemic compelled higher education institutions to adopt an alternative approach to learning. For mathematics education, supporting students through blended learning has become increasingly important as it will ensure students' learning is sustained in such a situation. This practical change has illuminated how cloud technology can be employed in mathematics education to improve the instruction process. Using a cloud classroom blended learning-instructional framework for integrating educational cloud tools into mathematics teaching, this study analysed students' metacognitive ability in mathematical problem-solving. A pre-experimental research using a one group pre-test-post-test design was conducted. The sample comprised sixty undergraduate mathematics students enrolled on a Numerical Analysis course. Sample selection was carried out using a simple random sampling technique. Data were analysed using a descriptive analysis, parametric tests, and n-gain. The results revealed that, firstly, the majority of students (90%) achieved a post-test score of metacognitive ability in mathematical problem-solving -- this exceeded the 60% threshold. Secondly, students' post-test score of metacognitive ability in mathematical problem-solving (mean score of 82.14, S.D. = 8.29) increased significantly compared to the pre-test score (mean score of 73.54, S.D. = 8.38). Further, the effect size was large. Thirdly, there was an enhancement of metacognitive ability in the mathematical problem-solving of students with a mean n-gain of 0.34, which is in the moderate category. Thus, cloud classroom blended learning significantly improved metacognitive ability in mathematical problem-solving among undergraduate students. Educators can apply the results to assess mathematics learning in order to improve the quality of metacognitive ability in mathematical problem-solving.