Creativity · MCP — server

MCP Apache Druid Server

Apache Druid is an open-source real-time analytics database built for streaming ingestion and sub-second OLAP. The MCP Druid Server lets Claude and other MCP clients discover datasources, inspect dimensions and metrics, and run Druid SQL — great for clickstream, ad-tech, and operational dashboards.

MCP facts

Kind: server
Ecosystem: anthropic-mcp
Language: Python
Transports: stdio

Capabilities

Tools: list_datasources, describe_datasource, run_druid_sql
Supports Druid's HTTP SQL endpoint
Basic auth or JWT pass-through

Install

pip install mcp-druid

Configuration

{
  "mcpServers": {
    "druid": {
      "command": "uvx",
      "args": ["mcp-druid"],
      "env": {
        "DRUID_BROKER_URL": "https://druid-broker.example.com",
        "DRUID_USERNAME": "claude",
        "DRUID_PASSWORD": "<password>"
      }
    }
  }
}

Frequently asked questions

Why Druid for LLM-driven analytics?

Druid's pre-aggregated, columnar storage makes it ideal for interactive question-answering over huge event datasets — the model can explore freely without bottlenecking the database.

Does it support native JSON queries?

This server focuses on Druid SQL for simplicity; native JSON is possible via a raw_query escape hatch in some builds.

Is there an official Druid server?

No official Apache Druid MCP server exists yet — this is a community package. Check the MCP servers registry for the most-maintained fork.

Sources

Apache Druid Documentation — accessed 2026-04-20
Model Context Protocol — accessed 2026-04-20