snowflake-mcp-setup

Snowflake MCP Server Setup

This skill guides you through connecting Cursor to a Snowflake-managed MCP server, enabling AI agents to securely interact with Snowflake data and services — including Cortex Search, Cortex Analyst, SQL execution, Cortex Agents, and custom tools (UDFs/stored procedures) — directly from the IDE.

Setup

1. Create a Programmatic Access Token (PAT)

Authentication uses a Programmatic Access Token. Generate one in Snowsight under Settings → Authentication → Programmatic Access Tokens. Use the least-privileged role that has USAGE on your MCP server and its tools.

Or with SQL:

ALTER USER <YOUR_USERNAME> ADD PROGRAMMATIC ACCESS TOKEN <PAT_NAME>;

See the SQL reference for full details.

2. Get Your MCP Server URL

The URL follows this format:

https://<account_url>/api/v2/databases/<database>/schemas/<schema>/mcp-servers/<server_name>

Your account URL is typically <orgname>-<account_name>.snowflakecomputing.com. Use hyphens (-) instead of underscores (_) in hostnames. Follow Snowflake docs to create an MCP server.

3. Configure mcp.json

The mcp.json is already configured to read from environment variables:

{
  "mcpServers": {
    "Snowflake": {
      "url": "${SNOWFLAKE_MCP_SERVER_URL}",
      "headers": {
        "Authorization": "Bearer ${SNOWFLAKE_PAT_TOKEN}"
      }
    }
  }
}

Before using the MCP server, make sure the following environment variables are set in your shell:

• SNOWFLAKE_MCP_SERVER_URL — Your full MCP server URL (see step 2).
• SNOWFLAKE_PAT_TOKEN — The Programmatic Access Token generated in step 1.