Databricks MCP Server

A Model Context Protocol (MCP) server for exploring Databricks Unity Catalog. Browse catalogs, schemas, tables, and query sample data.

Features

• 🔐 Secure U2M Authentication using Databricks CLI OAuth flow
• 📂 List Catalogs - Browse all available Unity Catalog catalogs
• 📁 List Schemas - Explore schemas within a catalog
• 📋 List Tables - View tables in a catalog.schema
• 📊 Table Details - Get column names, data types, and metadata
• 📝 Sample Data - Query sample rows from tables (5-50 rows)
• ♻️ Token Refresh - Gracefully handles token expiration

Installation

Using pip

pip install mareana-databricks-mcp-server

Using uvx

uvx mareana-databricks-mcp-server

From source

git clone https://github.com/mareana/databricks-mcp-server.git
cd databricks-mcp-server
pip install -e .

Prerequisites

1. Install Databricks CLI

# Using pip
pip install databricks-cli

# Or using homebrew (macOS)
brew tap databricks/tap
brew install databricks

2. Authenticate with Databricks

The server uses U2M (User-to-Machine) OAuth authentication via the Databricks CLI:

# Login to your Databricks workspace
databricks auth login --host https://your-workspace.cloud.databricks.com

# Or with a named profile
databricks auth login --host https://your-workspace.cloud.databricks.com --profile my_profile

3. Verify Authentication

databricks auth token
# Should output a valid access token

Configuration

Variable	Description	Default
DATABRICKS_PROFILE	Databricks CLI profile to use	DEFAULT

Example

# Use a specific profile
export DATABRICKS_PROFILE="my_workspace_profile"

Usage

Running the Server

With Python:

mareana-databricks-mcp-server

With uvx:

uvx mareana-databricks-mcp-server

From source:

python -m databricks_mcp_server.server

MCP Client Configuration

Add to your MCP client configuration (e.g., Claude Desktop):

{
  "mcpServers": {
    "databricks": {
      "command": "uvx",
      "args": ["mareana-databricks-mcp-server"],
      "env": {
        "DATABRICKS_PROFILE": "my_profile"
      }
    }
  }
}

Or using Python directly:

{
  "mcpServers": {
    "databricks": {
      "command": "python3",
      "args": ["-m", "databricks_mcp_server.server"],
      "env": {
        "DATABRICKS_PROFILE": "my_profile"
      }
    }
  }
}

Available Tools

who_am_i

Get the current authenticated Databricks user.

Returns: Username, display name, and email of the authenticated user.

Example Response:

{
  "username": "user@example.com",
  "display_name": "John Doe",
  "active": true,
  "emails": ["user@example.com"]
}

list_catalogs

List all available catalogs in Unity Catalog.

Returns: List of catalogs accessible to the current user.

Example Response:

{
  "total_count": 3,
  "catalogs": [
    {"name": "main", "owner": "admin@example.com"},
    {"name": "dev", "owner": "admin@example.com"},
    {"name": "staging", "owner": "admin@example.com"}
  ]
}

list_schemas

List all schemas in a specified catalog.

Parameter	Type	Required	Description
catalog_name	string	Yes	Name of the catalog

Example Response:

{
  "catalog": "main",
  "total_count": 2,
  "schemas": [
    {"name": "default", "full_name": "main.default"},
    {"name": "raw_data", "full_name": "main.raw_data"}
  ]
}

list_tables

List all tables in a specified catalog and schema.

Parameter	Type	Required	Description
catalog_name	string	Yes	Name of the catalog
schema_name	string	Yes	Name of the schema

Example Response:

{
  "catalog": "main",
  "schema": "raw_data",
  "total_count": 2,
  "tables": [
    {"name": "users", "full_name": "main.raw_data.users", "table_type": "MANAGED"},
    {"name": "events", "full_name": "main.raw_data.events", "table_type": "EXTERNAL"}
  ]
}

get_table_details

Get detailed information about a table including column names, data types, and metadata.

Parameter	Type	Required	Description
catalog_name	string	Yes	Name of the catalog
schema_name	string	Yes	Name of the schema
table_name	string	Yes	Name of the table

Example Response:

{
  "full_name": "main.raw_data.users",
  "table_type": "MANAGED",
  "data_source_format": "DELTA",
  "column_count": 3,
  "columns": [
    {"position": 1, "name": "id", "type": "LONG", "nullable": false},
    {"position": 2, "name": "name", "type": "STRING", "nullable": true},
    {"position": 3, "name": "created_at", "type": "TIMESTAMP", "nullable": true}
  ]
}

get_sample_data

Get sample data from a table.

Parameter	Type	Required	Description
catalog_name	string	Yes	Name of the catalog
schema_name	string	Yes	Name of the schema
table_name	string	Yes	Name of the table
limit	integer	No	Number of rows (min: 5, max: 50, default: 10)

Example Response:

{
  "table": "main.raw_data.users",
  "warehouse_used": "Starter Warehouse",
  "row_count": 3,
  "columns": ["id", "name", "created_at"],
  "data": [
    {"id": 1, "name": "Alice", "created_at": "2024-01-01T00:00:00Z"},
    {"id": 2, "name": "Bob", "created_at": "2024-01-02T00:00:00Z"},
    {"id": 3, "name": "Charlie", "created_at": "2024-01-03T00:00:00Z"}
  ]
}

execute_query

Execute an arbitrary SQL query on Databricks and return the results.

Parameter	Type	Required	Description
sql	string	Yes	The SQL query to execute
max_rows	integer	No	Maximum number of rows to return (default: 1000, max: 10000)

Example Response:

{
  "status": "success",
  "warehouse_used": "Starter Warehouse",
  "row_count": 2,
  "columns": [
    {"name": "matl_num", "type": "STRING"},
    {"name": "total_stock", "type": "DOUBLE"}
  ],
  "data": [
    {"matl_num": "9000082572", "total_stock": 150.5},
    {"matl_num": "9000009550", "total_stock": 42.0}
  ]
}

Token Expiration Handling

The server gracefully handles token expiration. If your token expires, you'll receive a helpful message:

❌ Authentication failed. Your token may have expired.
Please re-authenticate using:
  databricks auth login --profile my_profile

Simply run the suggested command to refresh your authentication.

Troubleshooting

"No SQL warehouses available"

Sample data queries require an active SQL warehouse. Contact your Databricks admin to ensure you have access to a warehouse.

"Permission denied"

You may not have access to the specified catalog/schema/table. Check your Unity Catalog permissions with your admin.

"Table not found"

Verify the catalog, schema, and table names are correct. Use list_catalogs, list_schemas, and list_tables to explore available objects.

License

MIT License - see LICENSE for details.

Author

Shashanka G - shashanka.g@mareana.com