Automating Grasshopper Plugin Development with AI

Most common web services today, such as Notion, Miro, and Google Suite, expose a

REST API, but connecting one to Grasshopper still means reading docs, writing an

HTTP client, building a component per endpoint, drawing icons, writing tests, and

packaging for Rhino 7-8 on Windows and Mac. This workshop dissects that workflow

and rebuilds it as a Claude Code plugin that can automate Grasshopper plugin

generation.

Part 1 - Learning. We use NotionConnect, a finished plugin, to understand the benefits

of connecting Grasshopper to a REST API, how it actually works, and the workflow it

takes to build one. Once we've broken that workflow down into repeatable chunks, we

compare the four ways you can teach an agent to run it using skills, slash commands,

MCP servers, and sub-agents. Using PleaseREST as an example, we show when to

reach for which: why C#/Grasshopper conventions belong in a skill, why "extract this

API's endpoints" belongs in an MCP server, why "retry the failed calls and escalate"

belongs in a sub-agent, and why the whole pipeline hangs off a single slash command.

By the end of Part 1, you'll have a method that applies to any repeatable workflow you

can take straight back to your company.

Part 2 - Building. Each participant can choose a real REST API and drive it through

PleaseREST and its 15 guided phases and 7 human review gates. This process will

develop it into a complete Grasshopper plugin with a typed client, components,

rendered icons, unit tests, docs, and .yak package. Rather than an autopilot that

guesses at code from documentation, PleaseREST discovers the API, validates every

request and response, scaffolds a buildable Visual Studio project, generates the code, and packages for Rhino 7/8 on Windows and Mac, pausing at every gate so nothing

irreversible happens without your approval. Participants leave with a finished plugin they

built, reviewed, and tested themselves, plus a reusable template for automating their

own repeatable AEC workflows. Prior C# experience is not required as the agent writes

the code. Your job is simply to design the workflow, steer it, and test the result.

Keywords:

workflow automation

AI agents

Claude Code

MCP

Grasshopper

REST API

Software & Versions (installed before the session)

Required

- Visual Studio Code

- Visual Studio 2022

- Claude Code - Pro/Max subscription because only it has claude code

- Python 3.10+ (for writing/running MCP servers)

- Rhino 7 or 8

Recommended

- Rhino 8 - Latest version, or newer than v8.23 if you want to run generating demo GH

files

- Cordyceps plugin - if you want to use a Grasshopper MCP for demo files

- GitHub account - only needed if you want to keep pushing your work, but you can do

without it