Snowflake

Connect AI assistants to Snowflake using the Cortex Agents MCP server, exposing Cortex Search, Cortex Analyst, and SQL execution over your account's data.

Type: Official

Source Type: Remote URL

Source: Snowflake Cortex Agents MCP

Description: Snowflake's Cortex Agents MCP server gives AI assistants governed access to data in your Snowflake account — Cortex Search for unstructured retrieval, Cortex Analyst for natural-language-to-SQL over semantic views, direct SQL execution, and custom UDFs/stored procedures. You define an MCP Server object in Snowflake, and Snowflake hosts the endpoint on your account with access enforced by Snowflake RBAC.

---

Setup

Step 1: Set up the MCP server in Snowflake

Follow Snowflake's Cortex Agents MCP guide to create the MCP Server object and a custom OAuth 2.0 client. When you create the OAuth security integration, use Natoma's callback URL as the redirect URI:

https://api.nova.cortex.snowflake.app/mcp/server/callback

The redirect URI must match exactly, or Snowflake will reject the OAuth flow.

From that setup, collect the database name, schema name, MCP server name, Client ID, and Client Secret.

Step 2: Add Snowflake in Natoma

In Natoma, an admin adds Snowflake under Apps and enters:

Field	Example
Snowflake account URL	example-id.snowflakecomputing.com
Snowflake database name	DATABASE_NAME
Snowflake schema name	SCHEMA_NAME
Name of the Snowflake MCP server	MCP_SERVER_NAME
Client ID	client_id
Client Secret	(from your OAuth client)

Step 3: Authorize your connection

When prompted while creating the connection in Natoma, click Authorize and sign in with the Snowflake user whose default role holds the privileges for the tools you exposed.

MCP sessions run under the user's default role only — secondary roles are not applied. Make sure the default role has exactly the privileges you want available, and a default warehouse is set.