Design Jewelry with Claude, ChatGPT or Gemini: RhinoArtisan MCP

Ask Claude, ChatGPT or Gemini for an oval two-carat ruby with a half-carat round diamond on each side. In RhinoArtisan 7.0, the stones land in your viewport.

Not a picture of a ring. Not a suggestion. The same pipeline as QuickGems: the right layers, the right materials, sized to their carat weight. Then you keep talking. What it weighs in 18k. What it costs, with certified diamonds quoted live on Nivoda. Whether it is ready to print, where the metal runs too thin, the sprues, the STL, the production report.

That is RhinoArtisan MCP. Claude, ChatGPT, Gemini or Copilot, plugged straight into the jewelry tools you already use.

What MCP actually is

MCP (Model Context Protocol) is the open standard the industry is converging on. Anthropic, OpenAI, Google and Microsoft all speak it. So does RhinoArtisan.

A local MCP server sits next to your install. The assistant you already pay for (Claude, ChatGPT, Gemini, Copilot) can call real RhinoArtisan tools: gems, settings, weights, pricing, manufacturing checks. The geometry it leaves is parametric and editable, as if you had modelled it yourself.

It also composes with everything that assistant already does. Your Claude Skills, your own prompts, the rest of your tools, now running on live data from the design in front of you.

Not a chatbot about jewelry. The tools themselves.

Most “AI for CAD” talks about the work. This one does it.

You can ask for the three-stone layout and watch it appear. You can ask for the 18k white weight and get a number from the model, not from a guess. You can ask whether the piece will print and get the thin spots, the sprues, the STL, the report.

Same engines. Same layers. Same materials you use by hand.

How it sits next to Assistant

RhinoArtisan 7.0 has two ways in, one idea.

Assistant lives inside the product. It already knows the piece you have open. Type “Change the main gemstone to Oval 1.50 cts” and the centre stone updates. Anything that changes the document is proposed first, with Apply and Cancel. It runs on frontier models (Anthropic, OpenAI, and through OpenRouter the rest: Mistral, DeepSeek). One key, every model. If RhinoArtisan runs, Assistant runs. No extra GPU.

MCP is the other door: the assistants you already use outside Rhino, wired into the same tools. If your studio lives in Claude, ChatGPT, Gemini or Copilot, you do not switch apps to get work done. You bring those conversations onto the model.

Use Assistant when you are inside the piece. Use MCP when the conversation started somewhere else, or when you want your own skills and prompts on top of real jewelry data.

A conversation that does not stop at the render

The useful part is that it does not end when the stones are placed.

Keep going, in the same thread:

• weight in 18k, or in platinum
• cost, with certified diamonds from live Nivoda stock
• print-ready or not, and where the metal is too thin
• sprues, STL, production report

That is the half of the workflow jewelers actually fight over: not the first sketch, the quote and the file that leaves the studio.

Who this is for

• Designers who already work in Claude or ChatGPT and want the answer to land in Rhino, not in a paragraph
• Manufacturers who need weight, thickness and a printable mesh from the same conversation that placed the stones
• Studios that have their own prompts and skills, and want those running on real models

RhinoArtisan 7.0 is still in beta. Features can still move before the final release. The idea will not: the AI reaches your real tools.

See the full 7.0 list on What’s New in RhinoArtisan 7.0.