Purpose: This is an MCP protocol-based Snowflake integration server that lets AI clients access the Snowflake database through standard tools, supporting read (default) and optional write operations, and is designed for enterprise proxy environments.
Main capabilities: Provides database operation tools (execute-snowflake-sql, execute-query, the latter enforcing read-only by default, allowing only SELECT, SHOW, DESCRIBE, EXPLAIN, and WITH statements without DML, and supporting limit/offset pagination and json/markdown/csv output), metadata discovery tools (get-connection-info, list-snowflake-warehouses, list-databases, list-schemas, list-tables, list-views, describe-table, query-view, export-schema), session in-memory note tools (add-note, delete-note, list-notes, get-note), and prompt templates (summarize-notes, analyze-snowflake-schema, generate-sql-query, troubleshoot-connection). It also supports resource subscriptions, resource templates, argument completion, and client-controlled log levels.
Usage: Can be installed with uvx (recommended), from source, or in development mode. Configuration requires copying .env.example to .env and filling in variables such as SNOWFLAKE_USER, SNOWFLAKE_PASSWORD, and SNOWFLAKE_ACCOUNT; MCP_READ_ONLY defaults to TRUE. Supports Docker Compose and VS Code MCP debugging (command uv, arguments run simple-snowflake-mcp).
Limitations and security: The security boundary relies on a least-privilege Snowflake role, not keyword filtering; read-only mode is controlled by the server configuration or MCP_READ_ONLY, and clients cannot relax it. It is recommended to set statement timeouts and rate limits; queries without an explicit LIMIT will be truncated with a notice. Snowflake errorsare not returned to the client verbatim.
Some installation commands, configuration examples, and Docker details are not fully provided.
