snowflake-mcp

The Snowflake MCP skill provides a robust bridge between Snowflake’s data infrastructure and Model Context Protocol (MCP) compatible clients, such as Clawdbot. By enabling native MCP communication with your Snowflake account, this skill allows users to treat Snowflake as a first-class tool provider, granting AI agents direct access to SQL execution, RAG pipelines, and advanced AI services. This is designed for data engineers, AI developers, and analytics teams who need to perform data-driven tasks, execute semantic queries, or orchestrate complex AI operations within a secure, governed Snowflake environment.

• Enables direct SYSTEM_EXECUTE_SQL operations for querying and data manipulation.

• Supports integration with Snowflake Cortex Search, enabling RAG (Retrieval-Augmented Generation) on unstructured documents.

• Provides interface capabilities for Cortex Analyst to perform natural language queries against semantic views and KPIs.

• Supports invocation of custom Cortex Agents for automated documentation retrieval and specialized domain reasoning.

• Facilitates the creation of custom generic tools via Snowflake stored procedures, including functions like automated email dispatch.

• Leverages existing Snowflake RBAC policies, ensuring all AI-driven data access adheres to corporate security standards.

• Prerequisites include a Snowflake account with ACCOUNTADMIN privileges and an active Programmatic Access Token (PAT) for authentication.

• Deployment involves executing CREATE OR REPLACE MCP SERVER SQL commands in Snowsight to define the specific toolset and service access.

• Configuration is managed via a standard mcp.json file, mapping the server URL and authorization headers for seamless client-side loading.

• Users must ensure network connectivity to their specific Snowflake account domain, adhering to regional URL formatting conventions.

• The tool supports various model types including SYSTEM_EXECUTE_SQL, CORTEX_SEARCH_SERVICE_QUERY, and CORTEX_ANALYST_MESSAGE, allowing for flexible configuration based on your specific workload requirements.