What's new: MCP Server (2026 release wave 2)
Business Central MCP server data tools in the 2026 release wave 2: four system tools (find tables, table relations, table schema, data query) let an agent build and run AL queries on BC data within the user's permissions. Also covers the new server features configuration UI and a global security toggle. Feature status is not stated.
- Introduction to MCP Server and agenda
- Query tools and four new system tools overview
- Demo setup and data analysis capabilities
- Agent workflow demonstration with customer data
- Query execution and compilation process
- Limitations and use cases for data tools
- New UX for MCP configurations and server features
- Security toggle and closing remarks
Overview
The video shows how the Business Central MCP server can support generic data analysis. The MCP host (Copilot Studio, VS Code or another) has an LLM construct queries, and Business Central compiles and runs them. Four system tools support this: find tables, table relations, table schema and a data query tool. The agent can search tables, see how they connect, read their fields and then run an AL query without needing an API.
The demo has an agent work through customer data in about 30 steps and 2 minutes, as an example of a long-running agent. The presenter notes that access depends on the user’s permissions and that the tools carry the limits of AL queries, including limits on returned data and query runtime. The video also shows a new configuration UI that separates server features (data tool, APIs) and a toggle on the Copilot and Agent capabilities page to turn the MCP server feature on or off.
Key points
- Four new system tools support data analysis: find tables, table relations, table schema and data query. Enabling the data tool server feature gives you these tools.
- The MCP host (Copilot Studio, VS Code or other) has the LLM construct the queries, and Business Central compiles and runs them, so APIs are not needed.
- What data the agent can reach depends on the user’s existing permissions, the tables the user can access, and how well the agent writes AL queries.
- In the demo, the agent searches for tables, reads their schemas and table relations, and can use the Microsoft Docs MCP to learn how to write AL queries.
- When a query fails to compile, the agent gets the messages straight from the AL compiler and can often fix the query itself.
- Limits: the tool can do only what AL queries can do, and queries have limits on how much data comes back and how long they run.
- The new server features box in MCP configurations shows which system tools each enabled feature, such as the data tool, opens up.
Features
| Feature | Status | At |
|---|---|---|
| Find tables tool | generally available (roadmap Run data queries with MCP Server), demoed | 1:16 |
| Table relations tool | generally available (roadmap Run data queries with MCP Server), demoed | 1:16 |
| Table schema tool | generally available (roadmap Run data queries with MCP Server), demoed | 1:16 |
| Data query tool | generally available (roadmap Run data queries with MCP Server), demoed | 1:16 |
| Generic data analysis for agents | status not stated, demoed | 1:44 |
| Server features configuration UI | status not stated, demoed | 6:04 |
| MCP server security toggle | status not stated | 7:14 |
A status with a roadmap link comes from the Microsoft 365 roadmap feature this part of the video covers (matched by Haiku; links Opus dropped are not used); other statuses need a status word in the video itself.
AL objects mentioned
As heard in the captions. A name that matches one object page by exact type and name links to it; the others stay as named.
- query “AL query” at 2:17
Not found in BC28-30: query “AL query”.
Quotes
- 0:50 “have your MCP host whatever that is copilot studio vs code or other have those LLMs construct your queries then we will compile and”
- 2:01 “with the data query tools uh it completely depends on the permissions that you already have the tables that you have access to in”
- 2:46 “It takes 30 steps. It takes 2 minutes. It’s a good example of a longunning agent and it comes up with a very good”
- 4:17 “it also looked at table relations essentially understanding how is one table connected to other tables in business central because that’s how you write”
- 4:38 “if it compiles a query and it fails, the compilation messages you get is directly from our compiler, right?”
- 5:05 “since this data is based on queries that also means that this tool can basically has the power of queries and the limitations of”
- 5:48 “there are also limitations in queries in business central on how much data can be returned and for how long the query can run”
Disclaimers in the video
- 5:36 subject-to-change: If you have data questions that will kind of result in a query that is not supported in the query or even optimal then of course that will not work.
- 5:48 subject-to-change: there are also limitations in queries in business central on how much data can be returned and for how long the query can run
Presenters (as heard): Kenny Pontabidan, Pushad Dvidi, Porshad.
