Many attempts are aiming to provide social robots with human-like cognitive abilities such as memory, language comprehension and visual and spatial processing. Cognitive architectures in combination with Large Language Models (LLMs) have the potential to act as basic components for such a system. We demonstrate the use of an Adaptive Control of Thought-Rational (ACT-R) model in combination with an LLM to store experiences from human-robot interaction (HRI) in the declarative memory of the cognitive architecture for a humanoid social robot. These experiences can be retrieved from memory as associated recollections and used for the robot’s actions and for prompt augmentation of the LLM. This type of memory also allows the creation, storage and updating of person models from interactions with different people, which enables the robot to get to know these people better through temporally unrelated interactions and to respond to them individually.