Dataverse MCP SQL has a 20-record result limit
Diesen Beitrag auf Deutsch lesen
McpExecuteSqlQuery in the tested Dataverse MCP setup limits returned rows to 20, while supporting focused filtering, joins and basic aggregates.
Original article from July 28, 2026: The Adventures of Dataverse MCP Server: The Mystery of the 20-Record Limit, by Rachit Garg, on Brewing Thought. Read the original
Source may be outdated. The original article is more than 60 days old. Dates and availability refer to the state of the source.
This is our own summary, not a republication or full translation.
TL;DR
The tested McpExecuteSqlQuery implementation exposes a top parameter with a default and enforced maximum of 20 returned records. Dataverse MCP SQL can still use SELECT, WHERE, ORDER BY, GROUP BY, JOIN and basic aggregates, but it is intended for focused AI-agent context rather than bulk retrieval or full analytics.
Why it matters
- Makers: Design agent queries to return a small, relevant result set and use supported filtering or aggregates where appropriate.
- Admins/IT: Test queries against Dynamics 365 Finance and Operations virtual tables because their provider can behave differently from native Dataverse tables.
- BI/Data teams: Keep bulk exports, frequently refreshed dashboards and large-scale retrieval in data and analytics tools rather than MCP SQL.
Frequently asked questions
Why does Dataverse McpExecuteSqlQuery return only 20 records?
The tested implementation enforces 20 as the maximum returned result size. Its Custom API metadata also exposes a top parameter with a default of 20.
Which SQL operations does Dataverse MCP SQL support for Copilot Studio agents?
It supports SELECT, TOP, WHERE, ORDER BY, GROUP BY, JOIN, COUNT, SUM, AVG, MIN and MAX. It excludes constructs such as subqueries, HAVING, DISTINCT and CTEs.
Do Dynamics 365 Finance and Operations virtual tables work the same with Dataverse MCP SQL?
No, virtual table results also depend on the underlying provider. Aggregate scenarios tested against F&O virtual entities did not always behave like native Dataverse tables.