NewsletterSubmit a tool

Official · by Snowflake-Labs

Snowflake MCP server

The Snowflake MCP server lets an AI assistant query structured Snowflake data and retrieve across structured and unstructured data.

  • Runs locally
  • Apache-2.0 licence
  • ★ 300 on GitHub
  • Verified against its source 19 Oct 2026
View source

What the Snowflake MCP server does

The server exposes tools such as Cortex Analyst, which queries structured data in Snowflake using semantic modeling, and Cortex Agent, which orchestrates retrieval across structured and unstructured data. It can also run LLM-generated SQL, subject to user-configured permissions. It runs locally and connects to Snowflake using the Snowflake Python Connector, supporting username/password, key pair, OAuth, and SSO authentication.

Tools it gives your assistant

3 tools or toolsets, as its documentation lists them.

  • Cortex Analysttoolset Queries structured data in Snowflake via semantic modeling
  • Cortex Agenttoolset
  • sql-executiontoolset Runs LLM-generated SQL governed by user-configured permissions

How to install it

Run it locally

With uv (uvx)

uvx snowflake-labs-mcp --service-config-file /tools_config.yaml --connection-name "default"

What you need

In the Snowflake-Labs documentation's own words:

  • “Yes. Pass it to the CLI flag --password or set as environment variable SNOWFLAKE_PASSWORD”
  • “The server supports all Snowflake Python Connector authentication methods including username/password, key pairs, OAuth, and SSO”
  • “If using a Programmatic Access Tokens, note that they do not evaluate secondary roles”

Commands and settings change between versions. Check the official instructions before you install.