Background: Online learning is becoming increasingly popular among learners. To enhance the effectiveness of online learning, researchers have embedded an affective pedagogical agent (PA) on the computer screen to help regulate learners' emotions and support their learning. However, previous research has paid little attention to the effects of affective PA on learners' metacognitive processing. It has also been shown that learners often do not spontaneously engage in metacognitive regulation behaviors without intervention. This study addresses this gap by examining whether affective PA provides metacognitive prompts can lead to better learning outcomes. Objectives: To explore the effects of aligning instructional design methods (e.g., adding an affective PA to computer screen) with appropriate instructional prompts (e.g., metacognitive prompts) on students' affective processing, cognitive processing, metacognitive processing, and learning performance in an online learning scenario. Methods: In a 2 (Affective PA: Affective PA vs. Neutral PA) × 2 (Metacognitive prompts: With vs. Without) between-subjects design, one hundred and twenty-four college students learned about 10 psychological concepts delivered either by an affective PA or a neutral PA. All participants' physiological arousal was recorded throughout the learning process, followed by a series of learning outcomes. More specifically, learners' physiological arousal was recorded via heart rate and skin conductance sensors, and their self-reported emotions and learning outcomes were measured using validated questionnaires. Results: The results showed that (a) Learning with an affective PA had higher physiological arousal, germane cognitive load, and transfer performance than learning with a neutral PA; (b) Metacognitive prompts improved transfer performance; (c) A neutral PA inducing learners to engage in metacognitive processing led to better learning outcomes, while affective PA providing metacognitive prompts did not facilitate learning. Conclusions: The findings demonstrated the positive role of affective PA and metacognitive prompts in promoting learning. The results of this experiment identified the interplay effects between affective PA and metacognitive prompts, which provided new perspectives for understanding the role of combining instructional design and learners' active processing to facilitate learning.