GraphAgent is an automated agent pipeline designed to manage explicit graphic dependencies and implicit semantic interdependencies to suit real-world data scenarios involving predictive tasks (e.g., node classification) and generative tasks (e.g., text generation). It consists of three key components: a graphical generation agent that constructs knowledge graphs reflecting complex semantic dependencies; a planning agent that interprets various user queries and formulates corresponding tasks; and an execution agent that efficiently performs planned tasks while automating tool matching and invocation. GraphAgent reveals intricate relational information and data semantic dependencies by integrating language and graphical language models.