OpenEdge MCP Server How To Guide

his video is a demonstration of how you can connect your Progress OpenEdge data to an AI session that's connected to our Sports 2020 database. For this demo, I'm going to use our MCP server connected through Visual Studio Code with the GitHub Copilot plugin.

For additional information please visit: https://www.progress.com/openedge/mcp-server

OpenEdge MCP Server Demo Transcript
Hello, my name is Dan Mitchell.

Today I’m going to give you a demonstration of how you can connect your OpenEdge data to an AI session that’s connected to our Sports 2020 database.

For this demo, I’m going to use our MCP server connected through Visual Studio Code with the GitHub Copilot plugin. But truthfully, any agent that can talk to an MCP—including Progress Agentic RAG—should work.

In this demonstration, I’ll be an accounts receivable clerk who wants to do some account research. So I’m sitting down at my chat session and I say I want to do some research on accounts receivable. What data should I be looking at?

The system translates that into a request to determine which data might be relevant. It looks at customers, invoices, orders, and other related data. Then it comes back with an evaluation and suggests which fields may be useful.

As it works through this, you can see its reasoning in Visual Studio Code. Occasionally it may go down the wrong path and show an internal error, but that’s not something you need to worry about. What matters is the final answer.

In this case, it recommends looking at invoices, customers, bill-to information, sales reps, and orders. It also suggests looking at invoices where the total amount is greater than the amount paid.

What it has done is use AI—in this case Claude Opus—to interpret the data and help guide how to perform the analysis.

For this scenario, I’m particularly interested in customers, so I ask: which customers have the highest balance?

The system takes that request, translates it, and retrieves data from the customer table using a business entity exposed through a REST service. As it processes the request, it generates Python scripts to render the results.

It returns a result showing that Hook, Line and Sinker has the highest balance. It also provides additional analysis, highlighting other insights to help explore the data further.

I decide to drill deeper into that customer and ask for more information. The system retrieves data from the customer table and also examines related orders, invoices, and line items to provide a more complete analysis.

It returns detailed information about the customer and highlights key observations. For example, all four orders are still in order status, and there are no invoices.

Next, I look at those orders and ask about the largest one. The system queries the order table and identifies the largest order, showing the associated products. It highlights that snowshoes make up a significant portion of the order value.

At this point, I want to understand whether this high balance is something I should be concerned about. I ask the AI to evaluate that.

It analyzes the data and points out some concerns: the customer has a balance with no invoices, and orders have been sitting for three years. It suggests possible issues and recommends next steps, such as following up.

I then ask for contact information for James Haggerty, intentionally misspelling the name. The AI correctly identifies the person and provides contact details, noting that only a phone number is available.

To get a broader view, I ask if this is my worst customer. The system evaluates multiple data points across customers and determines that while Hook, Line and Sinker has the highest balance, it is not necessarily the worst customer based on other metrics.

It provides a more complete picture, indicating that other customers may be at higher risk based on credit limits and utilization.

From here, I can continue exploring. For example, I can shift focus to invoices and ask which customers have the most unpaid invoices.

The system analyzes invoice data and identifies customers with the highest number of unpaid invoices, as well as those with the largest outstanding dollar amounts.

At this point, I’ve gathered enough information to understand how AI can help guide accounts receivable analysis and decision-making.

That concludes the demonstration. Thank you for your time.