WS1
Agentic AI Design Orchestration using MCP
Learning Objectives
Participants will be introduced to Model Context Protocol (MCP) for agentic automation of the design process inside Claude Desktop.
The workshop will center on real-world-inspired architectural geometry as a case study. Participants will build their first orchestration hub to control the parametric exploration of architectural design options through chatting with an AI agent.
The Rhino/Grasshopper parametric model will be connected to external Python functions and Excel databases for performance evaluation, real-time visualizations, and building custom apps.
The final part of the workshop will focus on participants' projects and/or thesis to help define custom MCPs based on their specific needs.
Workshop description
The current status quo of large language models (LLM) and emerging trends in agentic orchestration provide new opportunities for early design investigation, geometry optioning, and performance evaluation. Although the LLM itself deals with uncertainties, these new paradigms enable opportunities for connecting and using actual developed and reviewed tools (like Grasshopper files or Python modules) while having the LLM as an assistant to chat with and connect different pieces.
Based on Walter P Moore's (WPM) well-known reputation in the design of long-span roof structures, this workshop will demonstrate the use of MCP for configuring truss geometries based on a real-world project and its constraints. The workflow will demonstrate connecting parametric models to an MCP server so that we can navigate through design iterations via chat while updating the main geometry. Participants will then be able to adapt the workflow presented to an application of their choosing.
The workshop will combine theoretical presentations that provide a thorough understanding of agentic automation concepts driven by large language models with practical, hands-on sessions focused on constructing various components of the orchestration hub. The program will conclude with a session where participants present their projects and explore how to implement relevant protocols and develop custom tools to address their specific design challenges.
Schedule
Day One – Morning:
Introduction on Walter P Moore AI initiatives and workshop objectives
AI in architectural design, LLM advances and agentic AI concepts
MCP fundamentals, architecture and real-world examples
Day One – Afternoon:
Setup first MCP server and Claude Code integration
Configure HTTPS bridge for Rhino 8
First MCP-Rhino-Grasshopper integration test
Day Two – Morning:
Unified data structures for defining tools
Integrate tool pool: Grasshopper, Excel, OpenSeesPy
Test multi-tool orchestration
Explore potential and limitations
Day Two – Afternoon:
Custom project development / Individual debugging sessions
Future directions and resources / Presentations
Participant Prerequisites
Required skills: Advanced Rhino/Grasshopper knowledge and Python programming skills. Basic knowledge of data structures and API integration.
Required software: Rhinoceros 8, Visual Studio Code (or similar IDE), Microsoft Excel.
Required hardware: Personal laptop with Windows 10 or later.
Workshop Information
Workshop Leaders
Hossein Zargar, Walter P Moore
Hossein is a software developer exploring how data-driven computational design connects across disciplines through scalable application. His current role as an AI Developer II at Walter P Moore's New York office focuses on developing next-generation tools that leverage current advancements in machine learning and large language models for informed building design decisions. He holds a Ph.D. in Architectural Engineering from Penn State, where his research focused on bringing robotic fabrication and constructibility knowledge into early-stage computational design optimization.
Jared Friedman, Walter P Moore
Jared is a licensed architect and design technologist with a focus on materials, energy, and technology in the built environment. As Computational Product Manager at Walter P Moore in New York City, he leads initiatives within the structural engineering group that integrate data-driven processes and computational tools across a range of project types and scales. Much of his focus revolves around leveraging computational tooling for the tracking and reduction of embodied carbon in the built environment. Jared has taught computational design at Columbia University's GSAPP and has presented at numerous conferences and universities where he has shared his passion for innovation in the built environment.
Dan Reynolds, Walter P Moore
Dan is the AI Leader at Walter P Moore and a structural engineer with expertise in specialty structures. After delivering numerous complex and high-profile projects, Dan now leads a pioneering team of engineers and AI developers reimagining how buildings are conceived, designed, and delivered. The promise of a more sustainable and responsive built environment has inspired the team to pursue a design process that is more creative, intuitive and holistic. To support this vision, they are developing next-generation tools that empower designers with real-time insights, generative solutions, and a deep computational intuition. Their custom software applications and APIs bring data-driven decision-making directly into the heart of the design process, reshaping what’s possible at every scale. As AI capabilities continue to accelerate, the team is not only exploring new ways for humans to engage with design—naturally, meaningfully, and enjoyably—but also embracing the profound responsibility of guiding these tools toward ethical, transparent, and trusted use.