Install guide and config at curatedmcp.com

Snowflake MCP: Query Your Data Warehouse Directly from Claude

The official Snowflake MCP server turns your data warehouse into a tool your AI agent can actually use. Instead of copy-pasting query results into Claude or manually checking schemas, your AI can now execute SQL, inspect tables, manage compute, and surface analytics—all in context, all in real time.

This is the integration analytics and data teams have been waiting for: direct, read-write access to Snowflake through Claude, Cursor, or Windsurf, without leaving your IDE or chat interface.

What It Does

Snowflake MCP gives AI agents native capabilities across your warehouse:

• Query execution: Run SQL and get structured results back into Claude instantly
• Schema inspection: Explore tables, views, and column metadata without switching tools
• Warehouse management: Scale compute up or down, check resource usage
• Task and stream control: Create, update, and monitor Snowflake tasks and streams
• Query performance: Pull query history, execution times, and bottleneck data
• External tables and shares: Access shared data and external sources directly
• BI integration: Feed analytics insights back into dashboards or reports

The practical outcome: ask Claude to "analyze our top customer segments from the past quarter" and it queries your actual warehouse, not a sample dataset. Or "which queries consumed the most credits this month?" and get real performance metrics in seconds.

How to Install

npx -y @snowflake/mcp

Then add to your Claude Desktop config:

{
  "mcpServers": {
    "snowflake-mcp": {
      "command": "npx -y @snowflake/mcp",
      "env": {
        "SNOWFLAKE_ACCOUNT": "your-account-id",
        "SNOWFLAKE_USER": "your-username",
        "SNOWFLAKE_PASSWORD": "your-password"
      }
    }
  }
}

Real-World Use Cases

• Ad-hoc analytics in your IDE: Ask Claude to run a cohort analysis or churn prediction query without opening Snowflake UI, then iterate on the results in code.
• Warehouse monitoring: Set up recurring checks in Cursor—"alert me if query costs exceed $500/hour" or "pull slow query logs and suggest optimizations."
• Data pipeline debugging: Inspect task history, check stream lag, validate external table schemas, all from your chat context when something breaks at 2 AM.

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.