MCP Server¶

The instant3Dhub MCP Server connects any MCP-compatible AI assistant to your 3DSpaces — querying structure, reading metadata, taking measurements, managing annotations, and more. It is the central component for Agentic 3D use cases.

Quick Start¶

1. Add the MCP server to your AI assistant (e.g Claude Code)

claude mcp add --transport http 3dspace https://<your-hub-instance>/mcp

2. Open a new 3DSpace

Open a new 3DSpace

3. Load a model

Add the following model to the space: <MODEL_URI>

Prerequisites¶

An instant3Dhub instance (version 3.13.x or later) running at your organization
Your instant3Dhub URL
An MCP-compatible AI assistant

Connect your client¶

The MCP endpoint for your instance:

https://<your-hub-instance>/mcp

Replace <your-hub-instance> with the address of your installation.

Claude Desktop¶

Open Settings → Connectors → Add custom connector. Enter the endpoint URL. Alternatively, edit your configuration file directly:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "instant3Dhub_MCP": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://<your-hub-instance>/mcp"
      ]
    }
  }
}

Claude Code¶

Run this command once in your terminal:

claude mcp add --transport http 3dspace https://<your-hub-instance>/mcp

The server is then available in every Claude Code session.

Other clients¶

Client	Documentation
Codex CLI	MCP servers in Codex
ChatGPT	Developer mode and MCP apps
Cursor	MCP configuration
VS Code (GitHub Copilot)	Add MCP servers
Windsurf	Cascade MCP integration
JetBrains IDEs	Model Context Protocol
Continue.dev	MCP deep dive

Authentication¶

The /mcp endpoint uses OAuth 2.0 as defined in the MCP authorization specification. Refer to the OAuth 2.0 documentation and your identity provider’s setup for details.

Contact your instant3Dhub operator for the credentials and flow configured for your installation.

What you can do¶

Once connected, your AI assistant can control 3DSpaces and the models inside them:

Query and navigate the product structure
Read and write node properties (visibility, color, metadata)
Transform nodes (position, rotation, scale)
Create, update, and remove annotations
Define and modify clip planes
Measure distances, volumes, surface areas, and more

To browse all available tools with parameters and schemas, connect the MCP Inspector to your endpoint.

Example prompts¶

These prompts work once a model is loaded in your 3DSpace.

Connect & add a 3DSpace and Model

“Create a new 3DSpace and load the model at [model path or URL].”
“Load the model at urn:x-i3d:examples:step:W16.”
“Join 3DSpace [3DSpace ID].”

Explore the model

“What models are currently loaded in the 3DSpace?”
“Give me the top-level assembly structure as a list.”
“How many parts are in the scene?”

Find specific components

“Find all components whose name contains ‘bracket’.”
“Which parts have the material property set to ‘aluminum’?”
“Show me the children of the node at path [scene path].”

Analyze geometry

“What is the bounding volume of the selected component?”
“Measure the surface area of this part.”

Modify the scene

“Hide everything except the engine subassembly.”
“Set the color of the selected part to red.”
“Create an annotation at part named ‘ID13423’ with the text ‘Inspect weld seam’.”

Review workflow

“List all annotations in the scene with their positions and text.”
“Set a clip plane to show the cross-section at the midpoint of the model.”

What data reaches the LLM¶

Your AI assistant makes MCP calls to instant3Dhub directly — the LLM provider never connects to your network.

However, MCP tool results — node names, property values, metadata, measurements, scene structure — are sent to the LLM API as part of the conversation. 3D mesh geometry is not transmitted.

If your models contain classified metadata, confirm with your security team before connecting an AI assistant.