Large Language Models (LLMs) are increasingly being explored as intelligent assistants for engineering applications, where they can support users in interacting with complex software systems through natural language. In structural analysis and design, however, engineering workflows often involve multiple interconnected tools, specialized input formats, and verification requirements that make automation challenging. This thesis investigates the use of an LLM-based framework to assist finite element analysis workflows, using Code_Aster as the underlying analysis platform. A hybrid architecture is proposed in which the LLM interprets user requests and coordinates the workflow, while all engineering-critical tasks, including file generation, mesh and solver execution, and result extraction, are performed by deterministic Python tools.