In recent years, agent technologies driven by large language models (LLMs) have rapidly advanced, making significant breakthroughs in automated problem-solving. However, existing methods still struggle with generating complete backend projects, particularly when dealing with complex project structures and business requirements. To address this challenge, this paper proposes Auto-Backend, an agent-driven framework for automated backend project generation. AutoBackend is capable of parsing user requirements, automatically generating the necessary development documentation, and building backend projects based on these documents. The framework ultimately achieves automated deployment using containerization technology. Experimental results demonstrate that this approach excels in both executability and code quality, exhibiting strong robustness in generating complex projects.