Operational Ontology MCP server

This is a small example project that shows how to give an AI helper a shared model of your business data. It comes with a demo company that has two old order systems, and your AI can read across both of them and even cancel an order the right way. It is handy if you are curious how AI can safely make changes in business systems, not just answer questions.

What this MCP server does

You ask your AI a question about the demo orders, like how many orders a customer has. Your AI uses this helper, and the helper reads from the shared business model that combines two old order systems. For actions, you ask your AI to cancel an order, and the helper checks the rules first. If the order already shipped, it refuses and tells your AI why. If it is allowed, it writes the change back to the original system and records what happened in an audit log.

Click to zoom

What you can do with it

• Search for orders by their details
• See all orders that belong to one customer
• Cancel an order if the rules allow it
• Assign an order to a person
• Add a note to an order
• Read the audit log of past action attempts
• Get a refusal message when an action breaks a business rule

Try asking your AI

• “How many unshipped orders does customer C-100 have?”
• “Cancel order O-55 because the customer changed their mind”
• “Show me all orders for customer C-100”
• “Read the audit log and tell me which cancellations were refused”

What it gives back to you

You get answers in plain chat: lists of orders, counts, and short summaries. When you ask for an action, you get either a confirmation that it was applied or a clear refusal with a code like SHIPPED_ORDER_CANNOT_BE_CANCELLED. The audit log comes back as a list of past attempts and their outcomes.

Before you start

Who it's for

People who want to see how AI agents can read and change business data safely, especially anyone curious about business rules, audit trails, and write-back to real systems.