MCP Server for Business Central: Connections and Permissions
An MCP server for Microsoft Dynamics 365 Business Central lets an AI assistant query business data, read records, and assist with follow-up tasks during an authorized session. Here is how to assess connection options, set the right action permissions, and use fenbs to track approved follow-up work without replacing the business system.
Updated 4 min read
An MCP server for Microsoft Dynamics 365 Business Central lets an AI assistant query customers, vendors, orders, inventory, and financial records through the Business Central API during an authorized work session. The assistant can retrieve specific records, summarize data, and help identify the next step without the user navigating through multiple Business Central pages. Tools like fenbs track the approved follow-up work and decisions rather than replacing the business system or copying sensitive records.
Business Central is a cloud ERP system used by small and medium-sized businesses for finance, operations, sales, and inventory. Its API is well-documented and supports both read and write access to most record types, which makes it a practical target for an MCP integration.
What a Business Central MCP server can expose
The Business Central API covers customers, vendors, items, sales orders, purchase orders, invoices, journal lines, and financial reporting. An MCP server wraps those endpoints as tools an AI assistant can call. Common read operations include looking up a customer's open orders, retrieving an invoice by number, checking an item's current stock level, and reading a vendor's payment terms.
Write operations include creating sales orders, posting invoices, updating item quantities, and adding journal lines. Those operations change live business data. Treat write operations as human-approved actions; the assistant helps draft or prepare the action, and a person reviews it before it is committed.
Assessing connection options
Business Central uses Microsoft Entra ID (formerly Azure Active Directory) for authentication. API access requires an app registration in your Microsoft 365 tenant with appropriate Business Central API permissions granted and admin consent applied. Check Microsoft's Business Central developer documentation for the current authentication requirements.
Search the MCP server directory for Business Central or Dynamics MCP tools. When evaluating one, confirm it supports the API version for your Business Central environment. Business Central's API version affects which record types and fields are available.
Setting action permissions before connecting
Create a dedicated app registration for the MCP connection with the minimum Business Central API permissions needed. Use read-only permissions as the starting point. The Business Central API permission model allows scoping to specific companies and record types.
Decide which operations require a person to confirm before the assistant takes them. Reading a customer record or checking an order status is low risk. Posting a journal entry, releasing a sales order, or applying a payment changes financial records. Those actions require a human review step.
Record the permission decisions as standing rules. A rule stating "the assistant reads Business Central records; a person reviews and confirms before any post or write operation" gives every team member and every AI session the same constraint without relying on verbal convention.
Keep Business Central access rules and follow-up work together
fenbs keeps Decisions and rules alongside tasks and Where we left off. Record the approved permission boundaries once, and track each piece of approved follow-up work as a task with its plan and status. Every connected AI assistant reads the rules before starting, so the approved boundaries hold across sessions, team members, and AI tools.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
Typical workflows with a Business Central connection
A useful starting workflow: ask the assistant to list all open sales orders for a specific customer, summarize the order amounts and expected delivery dates, and identify any overdue lines. That gives the account manager a current picture without navigating the sales order list manually.
Another common use: ask the assistant to find all invoices posted in the last 30 days for a vendor, check whether any are marked as disputed, and list the payment due dates. The assistant reads the data; the accounts payable team acts on what it finds.
For organizations running multiple Business Central companies in one environment, confirm which company the assistant is connected to before starting a session. Multi-company deployments include the company name in each API request, and a query against the wrong company returns data that looks valid but belongs to a different entity.
Keeping follow-up work separate from the business system
Business Central is the system of record for financial transactions. It is not the right place to track the conversation about those transactions or the team's next steps. When an AI session surfaces an issue, the next action belongs in a task tracker, not as an informal note inside the ERP system.
For general MCP setup guidance, including authentication and configuration patterns for API-based servers, see mcp-server-configuration. The azure-mcp-server article covers the Microsoft authentication approach that Business Central also uses.
How to do this with fenbs
- Create a free account and set up a project for the business function or team that uses Business Central.
- Record the approved access rules as Decisions and rules: which record types the assistant may read, which write operations require human confirmation, and who is responsible for reviewing the assistant's output.
- Connect your AI assistant over MCP from Integrations. Ask it to read the project context before querying Business Central so it knows the current rules and the pending follow-up items.
- When the assistant surfaces an issue or identifies a next step, create a task in fenbs for it rather than leaving it as a chat note. Assign it, add the context, and track it to completion.
- Before handing off to a colleague, save Where we left off with the open tasks, the records reviewed, and the next step. The colleague reads that note before their session rather than asking for a full re-briefing.