In this presentation, we introduce intelligent agents based on large language models. We will review the various architectures that have been proposed over time for connecting language models and prompting them for this purpose. The strengths and weaknesses of these architectures will also be examined. One of the popular architectures is the multi-agent architecture, whose different types will be analyzed, along with the conditions that can contribute to the success or failure of these agents. A key component in these agents is long-term memory, which is used to store valuable results obtained through high-level computations in previous stages. This important element will also be thoroughly discussed. Finally, examples of designing intelligent agents based on the mentioned concepts will be presented.