Power Platform Community Forum Thread Details

Hi Team,

I’m using Azure SQL Server as a data source in Microsoft Copilot Studio. When I ask natural language questions in the Copilot, it provides correct answers based on my database — which indicates it is firing SQL queries in the background.

However, I’m unable to trace or view the actual SQL queries that Copilot Studio is generating and executing against the Azure SQL Database.

I would like to know:

Where can I find the query logs or traces of the SQL statements Copilot sends to the database?
Is there any built-in option or diagnostic setting in Copilot Studio, Power Automate, or Azure SQL that allows capturing or monitoring those generated queries?
If not, what is the recommended way to audit or debug** the queries executed by Copilot Studio for transparency and troubleshooting.

I’m also attaching an activity snapshot for reference.

Any guidance or documentation link would be greatly appreciated.

Thanks & Regards,

Srujan V

Categories:

Bot analytics

At the moment, Copilot Studio doesn't expose the generated SQL query text when using Azure SQL as a knowledge source. The Activity pane shows the retrieved sources and citations, but not the SQL statements that were generated or executed.

For troubleshooting or auditing, your best options are on the Azure SQL side:

Enable Query Store to capture executed queries and their performance.

Use Azure SQL Auditing to log database activity.

Use Extended Events or SQL Profiler (where applicable) to capture SQL statements in real time.

Monitor Azure Monitor and Log Analytics if your database diagnostics are configured.

One thing to keep in mind is that Copilot Studio may not always generate a single SQL query. Depending on the scenario, it can translate the user's intent into one or more data retrieval operations before constructing the response.

Unfortunately, there isn't currently a built-in Copilot Studio setting that lets you view the exact SQL generated by the agent. If you need complete transparency into the executed SQL, database-side monitoring (Query Store, Auditing, or Extended Events) is the recommended approach.

Out of curiosity, are you using Azure SQL as structured knowledge (database knowledge source) or are you calling SQL through a Tool/Power Automate flow? The tracing options differ slightly between those two approaches.