Skip to main content
The Asset Discovery agent lets you explore your data assets through conversation. Instead of writing queries manually, you describe what you want to know — the agent retrieves relevant schema context and either answers directly or generates a SQL query for you.

​
What it does

  • Answers questions about table schemas, column names, and data types
  • Discovers relationships between tables using foreign key metadata
  • Generates SQL queries against your registered datasets
  • Provides data profile information for columns and tables

​
What you can ask

Here are examples of questions the agent handles well:
  • “What tables are in the orders schema?”
  • “What columns does the customer table have?”
  • “Show me all tables that reference product_id.”
  • “Write a query that joins orders and customers by customer ID.”
  • “Which columns in the sales table might contain PII?”

​
Prerequisites

Before using the Asset Discovery agent:
  1. Your data source must be connected and added to your workspace. See Data Sources.
  2. Datasets must be registered as assets. Go to Assets to confirm they appear.

​
Create an agent

  1. Go to the Agents tab in the left sidebar.
  2. Click Create Agent.
  3. Select Asset Discovery as the agent type.
  4. Select the data source, and add a name and description for the agent.
  5. Click Create to save.

​
Run an agent

  1. Go to the Agents tab and click the agent you created.
  2. Select the tables you want the agent to run on — you can select all tables or choose specific ones.
  3. Click Run. A chat window opens where you can ask questions and review results.
  4. You can create multiple runs for the same agent, each with a different table selection.

​
Output

The agent returns:
  • A natural language explanation
  • A SQL query (when applicable), formatted for your registered database dialect
  • References to the specific tables and columns it used

​
Limitations

  • The agent queries schema metadata and embeddings — it does not execute queries against live data unless you explicitly request SQL execution.
  • Results are based on the schema information registered at the time of the last sync.