The MCP server links your Omi account to your AI assistant. Once connected, you can query your catalog, your renders, and the progress of your modeling directly from your AI tool, in natural language.
⚠️ The connection is read-only: your AI assistant can view your data, but can't modify anything.
The Integration Manager role is required to configure the MCP server.
Connect the MCP server
In your Omi workspace, open the
Integrationssection in the left menu.Under MCP server, click
Configure.Select your client: Claude, ChatGPT, or Cursor.
Follow the connection method for your client (see below).
The connection always uses your Omi account, so the assistant has access to the same data you can access in Omi.
Claude
Claude
Two options:
In one command: run this command in your terminal.
claude mcp add --transport http omi https://mcp.omi.so/mcp
From the app: click
Add to Claude.A new tab opens in Claude, click
Add.
ChatGPT
ChatGPT
Go to Settings → Connectors.
Add a connector with this URL:
https://mcp.omi.so/mcp
You can also click Open Codex settings to configure it directly.
Cursor
Cursor
Add this to your MCP configuration file:
{
"mcpServers": {
"omi": {
"url": "https://mcp.omi.so/mcp"
}
}
}You can also click Add to Cursor.
What you can ask / Examples
Once connected, you can query your Omi data in natural language. The configuration window offers example suggestions, grouped by theme.
Getting started / organizations
"Which Omi organizations do I have access to?"
"List the teams under my organization."
Catalog
"How many collections, digital twins, and renders are in my catalog?"
"Show me the collections in my catalog with their thumbnails."
"Find the digital twin with SKU <SKU>."
"List the digital twins in the <collection> collection."
"Show me the variants of <digital twin>."
Renders
"Show me the latest renders for my organization."
"How many renders does <digital twin> have? Show them as a gallery."
"Which renders were created this week? Show the images."
Modeling
"Which digital twins are currently in modeling and what stage are they at?"
"Show me everything that has an open issue."
"What's the full history of <digital twin>? Has its model been approved?"
"List the 3D versions of <digital twin> and tell me which ones are validated."
Go further: combine your Omi data with your AI's capabilities
The MCP isn't just for viewing your data: your AI assistant can retrieve it and shape it to your needs.
For example, with Claude, you can ask:
"Create an Excel file listing all the digital twins in my organization, with these columns: the digital twin's preview (image), its name, and the number of renders that feature it."
The MCP server provides the data (names, images, render counts), and Claude assembles it into the Excel file. You get a ready-to-use export without leaving your AI tool.