What is MCP?
MCP is an open standard that allows AI applications to connect to external data and tools. To understand the difference, just think about how we have used a chatbot together with a modeling program so far. We can ask the AI, for example, to write a script for Rhino. The chatbot produces the code, but we are the ones who have to copy it into the program, run it and verify the result.
With an MCP connection some of these steps can happen directly. The software makes certain functions available and the AI assistant can call them to respond to the user’s request.
The passage, simplifying, is therefore from: prompt → response → user → software to: prompt → AI assistant → software tools.
This does not mean giving AI indiscriminate access to the program. It is the MCP integration that defines what information can be read and what operations can be performed. MCP essentially constitutes a common language through which the assistant can use external tools.
What can you do in Rhino today?
The case of Rhino makes this difference particularly evident. The Rhino MCP Platform, developed by McNeel, connects Rhino and Grasshopper to MCP-compatible assistants. The assistant can see the contents of files, create geometry, execute commands, write or edit scripts, and intervene on Grasshopper definitions.
The examples published by McNeel include very different operations: automatically distributing the elements of a model on the layers, producing variations of a geometry, building a Grasshopper definition through a natural language request or starting from an image to obtain a first parametric construction.
The difference compared to the normal use of ChatGPT or Claude is therefore concrete.
We can imagine, for example, asking: “Take the objects present in the model and organize them on different layers based on their type”. The assistant doesn’t just explain how to set the layers or suggest a script: he can directly use the Rhino tools made available to him to carry out the operation.
McNeel already presents Rhino MCP as compatible with several assistants, including Claude, OpenAI Codex, Gemini, GitHub Copilot, and locally run models.
What about other modeling programs?
Rhino represents one of the most structured examples today, but MCP is not a McNeel-specific technology. Experiments have also arisen around SketchUp that connect the program to AI assistants through MCP. In this case, however, it is necessary to distinguish between projects developed by the community and official integrations.
A significant signal came from Trimble itself: in May 2026 the company invited SketchUp extension developers to collaborate on the development of MCP projects for its Trimble AI Studio and Trimble Assistant systems. We are therefore not yet faced with a solution equivalent to the platform already available for Rhino, but the direction is similar: allowing AI assistants to interact with the tools used to build and modify models, instead of remaining confined to a chat window.
What changes for the designer?
The most interesting aspect, at least in the short term, is probably not asking AI to “design a building”. The applications on daily activities are much more concrete: organizing objects and layers, making repetitive modifications, producing series of variants, querying what is present in the model or using natural language procedures that today require commands, scripts or parametric definitions.
The prompt then begins to take on a different role. It is no longer just used to produce a text, an image or code: it can also become a way to recall and combine actions within work tools.
However, a fundamental point remains. Just because an assistant can perform an operation does not mean that it automatically knows whether that operation is correct from a design or technical perspective. It can build a geometry, modify dozens of elements or generate a Grasshopper definition. Establishing what to change, according to which criteria and whether the result is actually usable remains the designer’s responsibility.
Rather than eliminating human control, these integrations therefore shift the focus on which operations to entrust to AI, which tools to make them available and how to verify the results.
The weekly column “Architectural Prompting” is edited by experts Luciana Mastrolia, Giovanna Panucci and Andrea Tinazzo
>> If you are interested in these topics, also sign up to the free Linkedin Newsletter AI & Design for Technicians, we’ll talk about it here!
