drizzle-cube

Drizzle Cube Plugin for Claude Code

A Claude Code plugin that provides MCP (Model Context Protocol) tools for interacting with Drizzle Cube semantic layer APIs.

Features

This plugin registers two MCP servers for a complete Drizzle Cube integration:

AI-Powered Tools (from real Drizzle Cube MCP server)

Tool	Description
discover	Find relevant cubes by topic or natural language intent
validate	Validate queries with auto-corrections
load	Execute validated queries

REST API Tools (from plugin)

Tool	Description
drizzle_cube_meta	Fetch cube metadata (measures, dimensions, relationships)
drizzle_cube_dry_run	Validate a query and preview generated SQL
drizzle_cube_explain	Get query execution plan with performance analysis
drizzle_cube_load	Execute a query directly via REST API
drizzle_cube_batch	Execute multiple queries in parallel
drizzle_cube_config	View current configuration status

Architecture

┌─────────────────────┐     ┌──────────────────────────────────┐
│  Claude Code        │     │  Drizzle Cube Server             │
│                     │     │                                  │
│  MCP Servers:       │     │  /mcp (MCP endpoint)             │
│  - drizzle-cube-api ├────►│    - discover                    │
│    (HTTP to /mcp)   │     │    - validate                    │
│                     │     │    - load                        │
│  - drizzle-cube     │     │    + prompts, resources          │
│    (stdio plugin)   │     │                                  │
│    - meta           ├────►│  /cubejs-api/v1 (REST API)       │
│    - dry_run        │     │    - meta, dry-run, explain      │
│    - explain        │     │    - load, batch                 │
│    - load           │     │                                  │
│    - batch          │     └──────────────────────────────────┘
│    - config         │
└─────────────────────┘

Why two servers?

• The real Drizzle Cube MCP server provides AI-powered tools with prompts and resources
• The plugin provides direct REST API access for debugging, SQL preview, and batch queries

Installation

From GitHub

/plugin install cliftonc/drizzle-cube-plugin

Local Development

git clone https://github.com/cliftonc/drizzle-cube-plugin.git
cd drizzle-cube-plugin
npm install
npm run build
/plugin install ./drizzle-cube-plugin

Configuration

The plugin works out of the box with the demo server at https://try.drizzle-cube.dev.

Quick Start

No configuration needed for the demo server. Just install and use:

/dc-ask "show me productivity by employee"

Custom Server Configuration

To use your own Drizzle Cube server, configure both the MCP server URL and REST API URL:

1. MCP Server (in .mcp.json)

The plugin's .mcp.json registers the real MCP server. Update the URL to point to your server:

{
  "mcpServers": {
    "drizzle-cube-api": {
      "type": "url",
      "url": "https://your-server.com/mcp"
    },
    "drizzle-cube": {
      "command": "node",
      "args": ["${CLAUDE_PLUGIN_ROOT}/dist/index.js"]
    }
  }
}

2. REST API (in .drizzle-cube.json)

Create .drizzle-cube.json in your project directory:

{
  "serverUrl": "https://your-server.com",
  "apiToken": "your-optional-token"
}

Or create a global config at ~/.drizzle-cube/config.json.

3. Environment Variables (alternative)

export DRIZZLE_CUBE_SERVER_URL="https://your-server.com"
export DRIZZLE_CUBE_API_TOKEN="your-token"

Configuration Priority

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

Available Commands

Command	Description
/dc-setup	Configure server URL and verify connectivity
/dc-ask	Query data using natural language (AI-powered)
/dc-query	Build queries with full schema awareness
/dc-debug	Debug queries with SQL preview and execution plans
/dc-create-cube	Create new cube definitions
/dc-create-dashboard	Create dashboard configurations
/dc-add-chart	Add charts to existing dashboards

Usage Examples

Natural Language Queries (AI-powered)

Ask questions in plain English:

/dc-ask "show me total revenue by product category for last month"

This uses the discover → validate → load workflow from the real MCP server.

Direct Queries (schema-aware)

When you know the cube structure:

/dc-query

This uses drizzle_cube_meta to show available cubes, then drizzle_cube_load to execute.

Debug SQL

Preview the generated SQL without executing:

/dc-debug

This uses drizzle_cube_dry_run and drizzle_cube_explain for SQL preview and execution plans.

Check Configuration