If you've read our overview of what Rhino MCP makes possible, or our look at plugin-specific applications, you might be ready to actually set it up. The good news: Rhino MCP Platform is an official tool from McNeel — the makers of Rhino and Grasshopper — and it's free to install without touching any source code.

What You're Installing

The Rhino MCP Platform connects Rhino and Grasshopper to AI assistants including Claude, GitHub Copilot, Codex, Gemini, and local LLMs. Once installed, your assistant can see what's in your Rhino and Grasshopper files, create geometry, run commands, write and edit scripts, and build Grasshopper definitions directly.

Official site and starting point: mcneel.github.io/RhinoMCP

GitHub repository: github.com/mcneel/RhinoMCP

Step 1: Install the Rhino Plugin

The plugin is installed via Yak, Rhino's built-in package manager, and is available for both Rhino 8 and Rhino 9. Installation links are provided directly on McNeel's getting-started page, which detects and offers the right version:

mcneel.github.io/RhinoMCP/docs/getting-started

After installation, restart Rhino and run the MCPStart command to confirm the plugin has loaded correctly.

Step 2: Choose Your AI Assistant

Rhino MCP works with several AI clients, each with a dedicated setup page:

  • Claude Desktop — the easiest entry point, recommended if you're not sure where to start: setup guide
  • Claude Code — terminal-based, better suited if you already work in a CLI: setup guide
  • GitHub Copilot — via Agent mode in VS Code, pointed at the Rhino MCP router
  • OpenAI Codex and Gemini CLI — both terminal-based, configured by pointing their MCP config at the Rhino MCP router
  • Local LLMs via LM Studio — for users who want everything running on their own hardware, with no internet dependency: setup guide

For Claude Desktop specifically, setup is genuinely simple: install Claude Desktop, download the connector file (connector.mcpb) from McNeel's site, then in Claude Desktop go to Settings → Extensions → Advanced Settings → Install Extension, and select the downloaded file. That's the entire connection process — no manual JSON editing required.

Step 3: Try Your First Prompt

Once connected, McNeel's documentation includes a growing library of example prompts grouped by task type — useful for getting a feel for what phrasing works well before attempting more complex requests. This is linked from the main docs page: mcneel.github.io/RhinoMCP/docs

A Note on Local LLMs

If data privacy is a concern, or you'd rather not depend on a hosted AI service, the Local LLM route lets you run an open-weight model on your own machine via LM Studio. Worth going in with realistic expectations, though: local models are generally weaker at extended tool-use chains than hosted options like Claude, so expect more retries and the occasional dropped step on complex multi-tool tasks. McNeel's own guidance suggests Qwen3 as a reasonable starting model, with at least 16GB RAM.

Troubleshooting

If something doesn't connect as expected, McNeel maintains a dedicated troubleshooting page covering common setup issues, linked from the main documentation hub above.

Where to Go Next

With the platform installed, our earlier posts cover what to actually do with it: an overview of core geometry, layer, and block operations, and a deeper look at querying plugin data from VisualARQ and LandsDesign, plus clash detection and dimensioning. If you want to package up your own studio conventions, see our guide to building and distributing Rhino MCP Skills.

Back to blog