BusinessCentral

You run an MCP server that exposes Business Central data and operations to AI clients through JSON-RPC or REST. This server lets Claude, Copilot, and other AI tools access customers, items, orders, and more from Business Central in a secure, well-structured way, enabling powerful automated workflows and AI-assisted business processes.

How to use

Connect to the MCP server from your AI client using the provided HTTP API endpoint or run a local MCP instance for development. The HTTP endpoint hosts the server remotely and is suitable for production use, while the local stdio option is convenient for testing and iterative development on your workstation. You can list customers, fetch item data, retrieve orders, and create new customers through the exposed tools, then compose AI-assisted operations using those results.

Practical usage patterns include reading data from Business Central to inform AI decisions, performing lookups to enrich AI-generated responses, and pushing new customer records created by an AI assistant back into Business Central. Use the HTTP endpoint for production scenarios and the local stdio flow when developing and debugging the MCP server interactions.

How to install

Prerequisites you need before installing:

• Python 3.11+ and a compatible virtual environment tool
• Network access to the Azure-hosted MCP endpoint if you are using the remote server
• Optional: FastAPI and JSON-RPC client libraries for testing and development

python -m venv .venv
.venv\Scripts\activate  # Windows
source .venv/bin/activate  # macOS/Linux

pip install -r requirements.txt

Set up and validate environment and credentials if you plan to run the MCP locally for development. Then start the local MCP server and the REST API to mirror production behavior.

python -m bc_server.BusinessCentralMCP

uvicorn bc_server.http_server:app --reload --host 0.0.0.0 --port 8000

Additional sections

Security and authentication recommendations emphasize using Microsoft Entra ID (Azure AD) with OAuth2 in production. Ensure you request and store access tokens securely and apply least-privilege access for the Business Central APIs.

If you encounter API rate limits or gateway timeouts from Business Central, implement retry logic with exponential backoff to respect the API's limits and maintain robust operation of your MCP client.

Testing the deployed server can be done with a test suite that exercises the main endpoints: health checks, listing customers, creating customers, listing items, and listing sales orders. Use the REST API docs to understand available operations and response shapes.

Claude Desktop integration steps let you expose the MCP server to Claude via a local configuration. After adding the MCP server entry, restart Claude Desktop and verify that the newly registered tools are available to your AI workflows.

Available tools

get_customers

Lists customers from Business Central, enabling AI-driven retrieval of client data.

get_customer_details

Retrieves detailed information for a specific customer by ID.

get_items

Lists items from Business Central for lookup and catalog enrichment.

get_sales_orders

Retrieves sales orders to support AI-assisted order processing.

create_customer

Creates a new customer in Business Central from AI-generated data.