Claude Code Plugin

The Drizzle Cube plugin for Claude Code lets you query your semantic layer using natural language. Ask questions like “show me revenue by region” and get real results, validated queries, and dashboard configurations.

Installation

Install the plugin from the Claude Code marketplace:

claude /install-plugin github:cliftonc/drizzle-cube-plugin

Or add it manually to your Claude Code settings.

Configuration

The plugin needs to know where your Drizzle Cube API is running. Create a .drizzle-cube.json file in your project:

{
  "apiUrl": "http://localhost:4000/cubejs-api/v1",
  "apiToken": "your-optional-api-token"
}

Or use the setup command:

/dc-setup

Configuration Priority

The plugin reads configuration in this order:

1. .drizzle-cube.json in current project directory
2. ~/.drizzle-cube/config.json (global)
3. Environment variables (DRIZZLE_CUBE_API_URL, DRIZZLE_CUBE_API_TOKEN)

Available Commands

Command	Description
/dc-query	Build and execute semantic queries with natural language
/dc-debug	Debug a query - validate, show SQL, explain plan
/dc-setup	Configure API URL and authentication
/dc-create-cube	Generate a new cube definition from a Drizzle table
/dc-create-dashboard	Create a dashboard configuration
/dc-add-chart	Add a chart to an existing dashboard

MCP Tools

The plugin provides these tools via the Model Context Protocol:

Tool	Purpose
drizzle_cube_meta	Fetch all cubes, measures, dimensions, relationships
drizzle_cube_load	Execute a query and return results
drizzle_cube_dry_run	Validate query and preview generated SQL
drizzle_cube_explain	Get query execution plan for performance analysis
drizzle_cube_batch	Execute multiple queries in parallel
drizzle_cube_config	Check current configuration status

Usage Examples

Query with Natural Language

/dc-query show me average happiness by department for the last 3 months

Claude will:

1. Fetch available cubes from your API
2. Build the appropriate query
3. Validate it with a dry-run
4. Execute and show results

Debug a Query

/dc-debug {"measures": ["Sales.revenue"], "dimensions": ["Products.category"]}

Returns:

• Validation status
• Generated SQL
• Execution plan
• Performance recommendations

Create a Dashboard

/dc-create-dashboard Sales Overview

Claude will ask about your metrics and generate a complete DashboardConfig with KPIs, charts, and tables.

Example Conversation

You: Show me total employees and average salary by department

Claude: Uses drizzle_cube_meta to find available cubes, then builds query:

{
  measures: ['Employees.count', 'Employees.avgSalary'],
  dimensions: ['Departments.name']
}

Validates with drizzle_cube_dry_run, then executes with drizzle_cube_load:

Department	Total Employees	Average Salary
Engineering	45	$125,000
Sales	32	$85,000
Marketing	18	$78,000

Skills (Reference Documentation)

The plugin also includes reference skills that Claude can use for detailed information:

Skill	Description
dc-query-building	CubeQuery structure, filters, time dimensions
dc-analysis-config	AnalysisConfig for dashboards and persistence
dc-chart-config	Chart types, axis mapping, display options
dc-cube-definition	How to define cubes with measures and dimensions
dc-dashboard-config	Dashboard layout, portlets, filters

Requirements

• Claude Code - Install Claude Code
• Running Drizzle Cube API - Your semantic layer must be accessible
• Node.js 18+ - For the MCP server

Troubleshooting

”MCP tools not available”

Make sure:

1. The plugin is installed: claude /plugins
2. Your API is running and accessible
3. Configuration file exists with correct URL

”Authentication failed”

Check your API token in .drizzle-cube.json or environment variables.

”No cubes found”

Verify your Drizzle Cube server is running and has cubes registered. Test with:

curl http://localhost:4000/cubejs-api/v1/meta

Source Code

The plugin is open source:

• Repository: github.com/cliftonc/drizzle-cube-plugin
• Issues: Report bugs or request features