Beyond Words: Unlocking Intelligent AI Automation with MCP in D365 Business Central

-

Introduction

Model Context Protocol (MCP) is an open standard that makes it easier for applications to provide information to large language models (LLMs). You can think of MCP like a USB-C port for AI—just as USB-C offers a common way to connect devices, MCP provides a standard way to link AI models with different data sources and tools.

MCP uses a client-server setup, where different parts work together to share data:

  1. MCP Hosts – Programs like Claude Desktop, IDEs, or AI tools that need access to data.
  2. MCP Clients – Connect directly to servers to request and send data.
  3. MCP Servers – Small programs that provide specific functions through MCP.

For this proof of concept (PoC), I used two key parts of MCP:

  1. Resources – A way for MCP servers to share data, which clients can read and use to give better responses.
  2. Tools – Functions that let servers do tasks, such as pulling data from other systems or running commands.

I built both the MCP server and client inside Cursor, writing all the code myself with AI’s help; since I haven’t worked with Python in five years.

As part of this PoC, I created three tools that allow the LLM to get data from Business Central and provide useful insights.

References

Model Context Protocol IO - Introduction
Model Context Protocol IO - Examples

Configuration

Before setting up the MCP server, follow these initial steps:

  1. Install UV (if using Python) - This helps manage dependencies efficiently. If you’re using a different tool, install the relevant package manager.

  2. Create a new project - This will hold all the files related to your server.

  3. Set Up a Virtual Environment - This keeps your project isolated from system-wide dependencies.

Once your virtual environment is ready, follow these steps:

  1. Create a New Python File – This will contain the code for your MCP server.

  2. Set Up Environment Variables – Create a .env file to store sensitive credentials like API keys and connection details.

NEVER hardcode environment variables in your code!
Always use the .env file to keep your credentials safe.
Always add the .env file to your .gitignore file.

For reference, you can copy the template from the Quick Start Guide and modify it as needed.


Next, I'm loading all the environment variables into global variables that I'm going to be using through multiple methods.


I have created 3 helper functions which will -

  1. Get the OAUTH token from the relevant API.
  2. Generate headers using the OAuth token.
  3. Send a request to Business Central and return the response.

Next, I have 2 more functions which simply format the JSON strings into natural language that LLMs can understand better.


Now come the actual tools.
The get_customers method fetches all (capped at 50) customers using OData APIs and then using the format_customer helper function to convert it into natural language.

Adding the """Documentation Strings""" is very important as that is how the LLM knows which tool to use and what the parameters actually mean.


The get_customer_ledger_for_timeframe function fetches all the customer ledger entries for a particular period and formats it using the helper function.


The get_customer_ledger_for_customer  method fetches all the entries for a particular customer and formats it using the helper function.

Do note the use of @mcp.tool decorator at the top on each of the tools.


At the end, we have the usual block where we call the run method on our MCP instance.
The transport type "stdio" is used for process based communication.

Now go to Cursor settings or if you are using Claude Desktop then settings for that and then go to the MCP section.
The first time you open it, you should see something like the below screen where you'll have the option to add a new MCP Server or edit a configuration json file which has details of multiple MCP Servers.


Now, I'll add the details of my MCP Server.


Please ensure that the address is absolute and accessible to the program.
Also this might be a good time to ensure that the program you are using to develop this server (uv, node, python) are all properly configured and their PATHs available to the terminal.
You can confirm this by calling the --version on the respective tools.

Now, if you go back to settings, you should see your MCP Server listed with the appropriate status.


If you want to double check whether your configurations are correct, you can copy the "Command" mentioned under your server and try to run it directly in the terminal.
If it runs successfully, you should be good to go.

If you are doing any changes to the source code, do rerun the environment as I've observed simply restarting the server doesn't really change much.

At this point, you are all set to go!
You can open the LLM chat and start asking questions and playing around with the data.

Results 

Conclusion

In the past, LLMs could only respond based on their existing knowledge. Integrating custom data into an LLM’s knowledge base was a long and tedious process, and even when it was possible, the AI had very limited ways to interact with that data.

The Model Context Protocol (MCP) feels like the long-awaited solution to this problem. With MCP, we can now extend AI applications with custom data and logic, allowing LLMs not just to analyze information but also to perform meaningful actions. This opens the door to intelligent automations, where AI-driven decisions are based on real customer data and actions crafted by humans but executed by AI.

Microsoft has already incorporated MCP into Copilot Studio, and I believe we’ll soon see similar functionalities appear across Dynamics 365 products. This shift is set to bring exciting possibilities for both end users and consultants, making AI-powered solutions more dynamic and impactful than ever before.

Comments

Post a Comment

Popular posts from this blog

Leverage Postman for Streamlined API Testing in Finance and Operations

-
Image
Introduction Postman is a tool for testing APIs. It lets you send requests, automate tests, and share your work with your team.  We're going to connect Postman to our Finance and Operations environment so we can test out standard or custom APIs. I've already covered automating the testing in Postman in my blog here  so once the connections are in place you'll be good to go! Pre-requisites Postman Finance and Operations Environment Configuration We'll start with creating an App Registration in Azure Portal. Go to the Azure Portal (of the same tenant of your FnO Environment). Search for "App Registration" and click on "New Registration". Add a name for your new app and click on "Register." Once it is completed, you'll be taken to the Overview of the app. Here, click on the "Add a certificate or secret" under the "Client Credentials." Add an appropriate name and select the expiration date of the certificate as necessary...
Read more

Integration with Finance and Operations - From Basics (Part 2)

-
Image
Introduction Finance and Operations provides two major ways to interact with tables (or data entities) for external system using APIs; namely Custom Services and Data Entities. Data entities in D365 Finance and Operations simplify data management by grouping data from multiple tables. They make it easier to import, export, and integrate data with other systems. Custom services in D365 Finance and Operations allow developers to create web services for specific business needs. They enable external systems to interact with D365 F&O by exposing custom logic and operations. This helps in integrating and automating processes with other applications. In the previous blog, we saw how we can use Data Entities to create APIs. In this blog, we'll see how we can use Custom Services to create APIs. References Custom Service Development Exposing an X++ Class as a Data Contract Using Data Contracts Pre-requisites A Finance and Operations Environment Postman connected to FnO Enviroment (If you...
Read more

Getting started with development in D365 FnO - Create your own project

-
Image
Introduction Before we get started with a new project, there's some words that I want to introduce. The most basic element of a project is an object and it's metadata. An object  can be a table, a form, a report or even a EDT field. A group of such objects form  a project . A model  is used to logically group multiple projects together but in essence it holds multiple objects across multiple projects. It can also hold customization done on existing models.  A package  consists of multiple such models and this is what we finally deployed in a Finance and Operations environment using LCS. References: 1.  Models and packages 2. Application Explorer 3.  Properties Configuration The first thing we would need to do to get started with D365 Finance and Operations would be to create and deploy a Finance and Operations environment. I've demonstrated the same in my blog here . Now, once the virtual machine has been setup, we can create our first project here. Cr...
Read more