snowflake-mcp

The Snowflake MCP skill provides a robust bridge between Snowflake's managed infrastructure and MCP-compatible AI clients like Clawdbot. It is designed for data engineers, analysts, and developers who need to integrate governed Snowflake data directly into their AI workflows. By leveraging the native Snowflake MCP server specification, users can securely interface with database operations, semantic models, and AI services without requiring local middleware or custom infrastructure deployment.

• Native integration with Snowflake MCP Server specifications via Snowsight.

• Support for SYSTEM_EXECUTE_SQL to perform arbitrary queries directly on your data warehouse.

• Advanced RAG capabilities using CORTEX_SEARCH_SERVICE_QUERY for vector search across unstructured documentation and internal records.

• Natural language processing for KPIs and business metrics via CORTEX_ANALYST_MESSAGE.

• Capability to trigger autonomous tasks and workflows using CORTEX_AGENT_RUN.

• Generic procedure support for executing custom database functions, such as email dispatch or custom data transformation tasks.

• Prerequisites: Requires a Snowflake account with ACCOUNTADMIN privileges and a valid Programmatic Access Token (PAT).

• Setup: Involves defining the MCP server using SQL commands in a Snowsight worksheet and configuring the mcp.json file in your project root.

• Security: Inherits existing Snowflake RBAC (Role-Based Access Control) policies, ensuring that the AI agent's data access is as secure as your standard database connection.

• Connectivity: If you encounter SSL errors, ensure you use hyphens in your account name rather than underscores.

• Best practices: Always validate connectivity using the provided curl test before initializing deep agent workflows.

• Constraints: Ensure the warehouse specified in custom procedures is active and reachable during execution.