# Databricks Data Explorer MCP server

The Databricks Data Explorer MCP server enables LLMs to discover and retrieve structured data from Databricks environments governed by Unity Catalog through natural conversation. It provides tools to explore catalogs, discover tables, retrieve schema information, and run read-only queries using Unity Catalog without requiring direct interaction with the Databricks interface. Row limits, timeouts, and concurrency limits are applied.

# Uses

Use the Databricks Data Explorer MCP server to perform the following actions:

Discover catalogs, schemas, tables, and views in Unity Catalog
Search for tables by name or keyword across your data catalog
Retrieve column definitions and table metadata
Retrieve sample data to understand table structure and content
Execute bounded SELECT queries against a designated SQL Warehouse
Run read-only queries with automatic row limits and timeouts
Check execution status of long-running queries
Retrieve results from completed asynchronous queries

# Example prompts

What catalogs are available in Databricks?
Show me the schemas in the sales_data catalog.
What tables exist in the customer_analytics schema?
Find tables related to subscriptions.
What columns are in the customer_events table?
Show me a sample of data from the activity_logs table.
Query the top 100 customers by revenue from last quarter.
Get the status of my running query.

# Databricks Data Explorer MCP server tools

The Databricks Data Explorer MCP server provides the following tools:

Tool	Description
list_catalogs	Lists available Unity Catalog catalogs.
list_schemas	Lists schemas within a specified Unity Catalog catalog.
list_tables	Lists tables and views within a specified schema.
search_tables	Searches Unity Catalog tables by name or keyword.
get_table_schema	Retrieves column definitions and metadata for a specified table.
get_table_sample	Returns a bounded sample of rows from a table you specify.
execute_query	Executes a read-only SQL `SELECT` query against the configured SQL warehouse.
get_query_status	Retrieves execution status of a previously submitted asynchronous query.
get_query_results	Retrieves results of a completed asynchronous query.

# Install the Databricks Data Explorer MCP server

Complete the following steps to install a prebuilt MCP server to your project:

Go to AI Hub > MCP servers.

Click + Create MCP server.

Go to the Start with a template section and select the prebuilt MCP server you plan to use.

Click Use this template.

Provide a name for your MCP server in the MCP server name field.

Go to the Connections section and connect to your app account.

Select the connection type you plan to use for the MCP server template.

User's connection: MCP server tools perform actions based on the identity and permissions of the user who connects to the application. Users authenticate with their own credentials to execute the skill.
Your connection: This option uses the connection established by the recipe builder and follows the same principles as normal app connections.

Select your connection type

VERIFIED USER ACCESS AUTHENTICATION REQUIREMENTS

Only app connections that use OAuth 2.0 authorization code grant are available for user's connection. Refer to Verified user access for more information.

Complete the app-specific connection setup steps in the following section.

# Databricks connection setup

View Databricks connection setup steps

Complete the following steps to establish a connection with Databricks in Workato:

Enter a Connection name to identify your Databricks account in Workato.

Use the Location drop-down menu to select the location where you plan to store the connection.

Enter the name of your server in the Server hostname field. For example, example.cloud.databricks.com.

Enter the HTTP path. For example, sql/protocolv1/o/3957355953478232/0710-102114-bfnouzcv.

Enter the Port your server uses. The default is 443.

Use the Catalog drop-down menu to select the catalog you plan to use for this connection. The default value is hive_metastore.

Use the Schema drop-down menu to select the schema you plan to use for this connection. The default value is default.

Use the Database timezone drop-down menu to select the timezone you plan to use. Workato uses this to convert timestamps during reads and writes.

Use the Authentication type drop-down menu to select the authentication method you plan to use. Configure the following fields based on your selected method:

Click Connect to establish the connection.

# How to use Databricks Data Explorer MCP server tools

Refer to the following sections for detailed information on available tools:

# list_catalogs tool

The list_catalogs tool lists available Unity Catalog catalogs. Your LLM uses this tool to identify available top-level data groupings before locating schemas or tables.

Try asking:

What catalogs are available in Databricks?
Show me all the data catalogs I can access.
List the top-level catalogs in Unity Catalog.
What data groupings exist in my Databricks environment?

# list_schemas tool

The list_schemas tool lists schemas within a Unity Catalog catalog you specify. Your LLM uses this tool to identify namespaces within a known catalog before locating tables.

Try asking:

Show me the schemas in the sales_data catalog.
What schemas exist in the analytics catalog?
List all schemas in the customer_data catalog.
What namespaces are available in this catalog?

# list_tables tool

The list_tables tool lists tables and views within a schema you specify. Your LLM uses this tool to discover candidate tables before constructing a query.

Try asking:

What tables exist in the customer_analytics schema?
Show me all tables and views in the sales schema.
List the available tables in customer_data.raw_events.
What tables can I query in this schema?

# search_tables tool

The search_tables tool searches Unity Catalog tables by name or keyword. Your LLM uses this tool to find tables based on business concepts when table names are unknown.

Try asking:

Find tables related to subscriptions.
Search for tables containing customer activity data.
Look for tables about support tickets.
Find any tables with 'revenue' in the name.

# get_table_schema tool

The get_table_schema tool retrieves column definitions and metadata for a table you specify. Your LLM uses this tool to understand table structure, column types, and available descriptions before constructing a SQL query.

Try asking:

What columns are in the customer_events table?
Show me the schema for the sales.transactions table.
What's the structure of the activity_logs table?
Get the column definitions for the subscription_data table.

# get_table_sample tool

The get_table_sample tool returns a bounded sample of rows from a table you specify. Your LLM uses this tool to understand the shape or typical values of a dataset without retrieving the full dataset.

Try asking:

Show me a sample of data from the activity_logs table.
Get some example rows from the customer_events table.
What does the data in the transactions table look like?
Show me a few sample records from the subscriptions table.

# execute_query tool

The execute_query tool executes a read-only SQL SELECT query against the configured SQL warehouse with automatic row limits and timeouts. Your LLM uses this tool to retrieve structured data in response to business questions.

Try asking:

Query the top 100 customers by revenue from last quarter.
Get all active subscriptions created in the last 30 days.
Show me support tickets opened this week by priority.
Find the total sales by region for January 2026.

# get_query_status tool

The get_query_status tool retrieves execution status of a previously submitted asynchronous query. Your LLM uses this tool only when execute_query returns a query_id, indicating asynchronous execution.

Try asking:

Get the status of my running query.
Check if my query has completed.
What's the execution status of query jfsialsk32s?
Is my data retrieval query still running?

# get_query_results tool

The get_query_results tool retrieves results of a completed asynchronous query. Your LLM uses this tool after get_query_status indicates that execution has completed.

Try asking:

Get the results of my completed query.
Show me the data from query 9dsdlkdsl.
Retrieve the results of the customer analysis query.
What did my query return?

# Getting started

View and manage your MCP server tools in the Overview page Tools section. Tool management provides the following capabilities:

TOOLS MUST BE STARTED

Your LLM can only access active tools in your MCP server connector.

Last updated: 3/13/2026, 5:05:19 PM