Local LLMs

LM Studio is a desktop app for running open-weight models on your own machine. It speaks MCP, so once you point it at the Rhino server a local model can drive Rhino the same way a hosted one can.

If you’re choosing between assistants and aren’t sure, start with Claude Desktop; it’s the gentler entry point. Pick LM Studio when you’d rather keep everything on your own hardware or have data and privacy concerns.

Before you start

1. The Rhino-MCP-Platform plugin is installed in Rhino. See Getting Started if you haven’t done that yet.
2. LM Studio is installed, and you’ve downloaded a model with strong tool-use support. A good starting point is Qwen3, which drives MCP tools reliably. Plan on 16 GB of RAM as a minimum.
3. Crank up the context length. LM Studio defaults to a small context window, which the Rhino tool list alone will blow through. Open the model’s load settings and push the max context length as high as your machine will allow.

Wire up the Rhino server

1. In Rhino, run the MCPConnect command, choose the mcp.json tab. It gives the JSON needed for LM Studio to connect to the Rhino MCP.
2. In LM Studio, open the Program sidebar and click Install > Edit mcp.json (or open ~/.lmstudio/mcp.json directly).
3. Add an entry for the Rhino server, pasting the mcp.json tab
4. Save the file. LM Studio picks up the change without a restart; the rhino server should appear in the MCP tool list for any new chat.

Pick the Rhino version by changing the --default-version arg. Use 8 for Rhino 8, 9 for Rhino 9 WIP.

Try it out

Load your model in LM Studio, start a new chat with tool use enabled, and follow the prompts on the Try it out page.

Tips

• Tool use must be on. In the chat sidebar, make sure the rhino server is toggled on for the session. LM Studio defaults to asking before each tool call.
• Smaller models struggle. If the model keeps describing what it would do instead of calling tools, try a larger or more recent instruction-tuned model.