Security and data privacy FAQ for the Luminovo MCP

The MCP connects an AI assistant that your company chooses, such as Claude, to your Luminovo data. Through it, the assistant can read data, run workflows, and build apps inside Luminovo.

The MCP server runs in the same Microsoft Azure data center in Frankfurt, Germany, where your Luminovo data already lives. It uses the same users, roles, and permissions as the Luminovo web app. Your assistant can see and do only what the signed-in user can see and do.

The answers cover three parts of the MCP:

To connect an assistant, see Connect your AI assistant to Luminovo with the MCP.

The same data as the signed-in user, and nothing more. The MCP acts as that user: same tenant, same roles, same modules.

Every call from the MCP goes to the same Luminovo backend that the web app uses, with the same permission checks. A few backend operations are blocked for the MCP, so it can do a subset of what the user can do in the browser. Without a signed-in user, the MCP has no access at all.

Yes, in two ways:

No. The MCP has no AI model of its own. Your assistant is the AI model, and your company chooses and contracts it. You can use any assistant that supports remote MCP connectors, including an open-source model that you host yourself.

Two Luminovo features use their own AI model: apps that are built to call an AI model, and the Lumi assistant inside the Luminovo web app. Both use OpenAI models on Microsoft Azure in the European Union. Contact us concerning ITAR support.

Into your conversation with your AI provider. When your assistant reads a BOM, a price, or a calculation, that data goes to your provider in the same way as a document that you paste into the chat.

Retention, training, and processing location on the provider side are set by your contract with that provider. Luminovo is not a party to that contract. If your company needs a data processing agreement for that processing, it is between your company and your provider.

Yes. An AI assistant is needed only to build an app. After the app is built, your team opens and uses it in Luminovo like any other page, with no AI assistant.

If you do not want an AI assistant to see your data, you build the app on a test tenant that has none of your data. Luminovo then copies the finished app to your tenant, and your team uses it there. No AI assistant connects to your tenant at any step.

This applies to apps that do not call an AI model themselves. An app can be built to call an AI model, for example to summarize a BOM. That app sends the data it works on to an AI model on Microsoft Azure in the European Union.

Luminovo does not train or fine-tune AI models on your data. Your data is used only to answer your request.

Whether your AI provider trains on your conversations depends on your plan and contract with that provider. Check their terms.

Luminovo logs every MCP call. Each entry records the user, the tenant, the AI client, the tool or endpoint, the result, the duration, the time, and an error message if the call failed. It can also hold a short note from your assistant on why it made the call. The log does not store the data that a call returns.

For questions about your contract or your data processing agreement with Luminovo, contact your Customer Success Manager.

Email us at [email protected], or use the chat button in the bottom-right corner of your workspace. To report a security issue, write to [email protected].