Translation Agent is a machine translation demonstration project employing a reflective workflow. Developed by Andrew Ng and collaborators, it primarily leverages large language models (LLMs) for text translation and incorporates reflection on the translation results to provide improvement suggestions. The model offers high customizability, allowing adjustments to translation style, handling idioms and proper nouns, and optimization for specific regions or dialects. Despite being relatively immature, it has demonstrated the potential to be more competitive than traditional machine translation systems in certain instances.