Agentic AI for Conversational Aerodynamic Analysis and Optimization of Airfoils and Wings

Computational fluid dynamics (CFD) based aerodynamic analysis and optimization have shown great potential for accelerating aerospace designs. However, traditional workflows require numerous manual steps, including editing configuration files, performing GUI clicks, transferring files, and scripting, which create a high barrier to entry for many users. To streamline this process, this paper develops an AI agent that performs end-to-end geometry generation, meshing, CFD simulation and optimization, and design visualization, all driven by natural-language commands. At the core of this system is a Model Context Protocol (MCP) server that translates design intent, expressed in natural language, into executable run commands, manages background execution, and post-processes simulation results for interactive visualization. The MCP server exposes a suite of standardized functions, such as generating airfoil/wing geometries and meshes with prescribed profiles, mesh densities, and target yPlus values; running simulations with specified angles of attack, Mach numbers, and Reynolds numbers; performing design optimization under user-defined lift constraints; and visualizing flow fields (e.g., velocity, pressure, turbulence quantities), pressure distributions, and sectional airfoil geometries. The MCP server is connected to a large language model (LLM) client, which converts the user’s design intent into corresponding function arguments and orchestrates the overall workflow. This paper demonstrates the full conversational design pipeline, including geometry creation, meshing, simulation, optimization, and interactive visualization, using representative airfoil and wing cases. The AI agent is open source and built on the DAFoam framework, and it has the potential to significantly lower the barrier to aerodynamic analysis, design, and optimization.