About this project

mcp-server-snowflake is an independently maintained, Apache-2.0 licensed Model Context Protocol (MCP) server that lets an AI assistant work with Snowflake. The project states it is not affiliated with, sponsored by, endorsed by, or supported by Snowflake Inc., and notes that "Snowflake" and "Cortex" are Snowflake Inc. trademarks. It is distributed on PyPI and can be installed with uv or pip, run through a setup wizard (--init), pointed at a named Snowflake CLI connection (-c), or run in Docker. Capabilities described in the README Connection resolution: the server can read Snowflake CLI configuration from ~/.snowflake/connections.toml so an existing profile can be used without extra configuration. It supports switching profiles at runtime through snowflake_use_connection and listing them with snowflake_list_connections. Authentication options include programmatic access tokens or OAuth (token in the connection profile or SNOWFLAKE_TOKEN), RSA key-pair JWT (SNOWFLAKE_PRIVATE_KEY_PATH, optional passphrase), SSO via externalbrowser authenticator, and standard environment variables (account, user, password, warehouse, database, schema, role). Safety controls: a read-only mode (SNOWFLAKE_MCP_READONLY=1 or --readonly) disables DDL/DML mutation tools; destructive operations such as dropping databases, schemas or tables require an explicit confirm=True; and default row and timeout limits (SNOWFLAKE_MAX_ROWS=1000, SNOWFLAKE_QUERY_TIMEOUT=120) are intended to bound result sizes. Tool coverage: the README organizes 140 tools into 19 domain modules, including SQL queries and transactions (query, DML, cancel, query history, query plan, operator stats, begin/commit/rollback); databases, schemas, tables and views (list, describe, create, drop, clone, undrop, DDL, sampling, truncate); virtual warehouses (list, describe, create, drop, resume, suspend, resize, load history); stages and file operations; tasks and DAG pipelines; streams and change data capture; dynamic and Iceberg tables with external volumes and catalog integrations; Snowpipe ingestion; alerts and notifications; governance, RBAC, roles and users; network and password policies; SPCS compute pools and Streamlit; object tags and classifications; Horizon lineage plus masking and row access policies; programmability objects such as procedures, functions, secrets, sequences, integrations and event tables; Cortex AI helpers for completion, summarization, sentiment, answer extraction, translation, search, text embedding and analyst queries; and composite workflows such as health checks, table inspection and profiling, warehouse scale-and-execute, cloning recipes, exporting query results to a stage, account usage summaries and schema lineage discovery. Intended audience Data engineers, analysts and platform teams who want an AI assistant to inspect and operate Snowflake accounts through MCP, with guardrails available for read-only or confirmation-gated usage. Licensing, CI status, PyPI version and coverage are indicated by repository badges.