ANNOUNCEMENT & AVAILABILITY
Checked against four public primary Microsoft Learn pages, linked in the article and consulted on October 24, 2026. The discrepancy between general availability and preview remains. The sources consulted do not establish a specific rate for these queries. Activation in a specific environment was not verified, and no queries were executed or outcomes measured. The example and interpretation are identified as editorial proposals. The source date refers to the overview's stated last update, October 2, 2026; it does not establish a launch date.
Model Context Protocol (MCP) in Business Central overviewData queries from an AI client
Microsoft documents data queries through the Business Central online MCP server. MCP, short for Model Context Protocol, connects AI applications to tools and data. Compatible clients include Copilot Studio and Visual Studio Code with GitHub Copilot. Data queries are read-only. Business Central MCP overview.
For a project, the practical question is how to turn a business request into a verifiable answer. Evaluating this proposal requires defining what to measure, which access to allow and which commercial conditions apply.
Availability: official sources differ
The public pages consulted on October 24, 2026 give different indications of the feature's status.
The discrepancy remains unresolved. Activation in a specific tenant and regional rollout have not been verified. The documented scope is Business Central online; these sources do not establish MCP support for on-premises installations. Before planning a deployment, we propose checking the version and the feature's status in the chosen environment.
| Public reference | Stated status |
|---|---|
| Update 29.0 overview | General availability for data queries with MCP Server. |
| MCP overview | Data Query Tools (Preview). |
| Configuration guide | Data Query Tools (Preview), subject to change. |
Four tools to prepare and execute queries
The configuration guide says Data Query Tools is activated under Server Features and documents these four tools.
As an evaluation criterion, technical validation and business validation should be considered separately. A valid query might still use an incorrect period, duplicate amounts when joining data or misinterpret a metric. Checking these aspects forms part of our proposed evaluation.
| Tool | Documented function |
|---|---|
| bc_data_find_tables | Find tables. |
| bc_data_get_table_schema | Read fields. |
| bc_data_get_table_relations | Inspect relationships. |
| bc_data_query | Compile and validate an AL query; use returnData=true to execute it and retrieve rows. |
What to check in the configuration
Creating configurations requires MCP - Admin or equivalent permissions. Active controls the availability of that configuration and its tools. Configuration guide.
MCP acts with the signed-in user's permissions. Copilot Studio integration. In addition to read-only queries, the server can expose write operations through APIs, depending on its configuration. Overview.
Our proposed review for consultants and developers covers:
- Identify the account, company, environment and configuration that the integration will use.
- Check which tools appear and which operations they permit under that identity, including access that should be denied.
- If a configuration is deactivated, also check other available access before concluding that MCP access to the environment has been closed.
Costs: client requirements and unresolved charges
The integration guide requires a Copilot Studio user license and available Copilot Credits capacity. That client requirement does not, by itself, establish a specific rate for these Business Central queries. Copilot Studio prerequisites.
The four pages consulted do not establish a unit price or an effective billing start date for these queries. That documentary scope also does not establish that they are free. To prepare a budget, we propose clarifying:
- Which charges belong to the AI client and which, if any, apply to query execution in Business Central.
- Which unit generates consumption and how validation, errors, retries and retrieval of additional pages are treated.
- Which commercial conditions apply to the environment's availability status and which might change afterwards.
Proposed example: a verifiable sales query
This example is an editorial proposal for a test environment where the feature is available. It has not been executed and does not represent an observed result.
A bounded request would help evaluate both business interpretation and data retrieval. Before starting, the team should prepare a test dataset with an expected result and agree on the definition of net sales.
- Review the tables, relationships, filters, dates and currency used, including the treatment of credit memos.
- Reconcile the result against an independent calculation using the same dataset.
- Check whether more rows remain to be retrieved and record errors, retries and observed execution times.
- Repeat the evaluation with an account that has limited permissions to check the intended access boundaries.
In the test company, identify the five customers with the highest net sales last month. Before querying, clarify the dates, currency and treatment of credit memos. Identify the tables, fields and filters used, and state whether the result is incomplete.
Interpretation: where it could add value
Our interpretation is that the value for consultants and developers will depend on the work required to obtain a reliable answer and maintain it as questions change. Exploratory queries and recurring integrations may have different priorities: flexibility for the former; stability and repeatability for the latter.
Evaluation should compare the full effort of definition, configuration, verification and maintenance. This is an editorial criterion for assessing projects, not a demonstrated saving. The sources consulted support a description of capabilities, but do not quantify productivity, accuracy or performance improvements in a real deployment.
Created with AI assistance and reviewed by Nexo Community. Practical proposals and editorial interpretation are distinguished from documented facts.
Read the source ↗
