Snowflake MCP Server (Snowflake Labs)

Add to your client

Copy the config for your MCP client and paste it into its config file.

{
  "mcpServers": {
    "snowflake-mcp-server-snowflake-labs": {
      "command": "uvx",
      "args": [
        "snowflake-labs-mcp",
        "--service-config-file",
        "/absolute/path/to/tools_config.yaml",
        "--connection-name",
        "default"
      ],
      "env": {
        "SNOWFLAKE_USER": "<your-user>",
        "SNOWFLAKE_ACCOUNT": "<your-account-identifier>",
        "SNOWFLAKE_PASSWORD": "<your-password-or-pat>"
      }
    }
  }
}

Before you start

• Python with the uv package manager (run via `uvx snowflake-labs-mcp`), or Docker
• A Snowflake account and user with appropriate roles/warehouse access
• Snowflake credentials supplied via CLI flags or environment variables: SNOWFLAKE_ACCOUNT, SNOWFLAKE_USER, plus one of SNOWFLAKE_PASSWORD, key-pair (SNOWFLAKE_PRIVATE_KEY_FILE), OAuth, SSO, or MFA via --authenticator
• A `configuration.yaml` (service config) file declaring which Cortex services and managers to enable
• For Cortex Search/Analyst/Agent: pre-created service instances and semantic models in your Snowflake account

About Snowflake MCP Server (Snowflake Labs)

Tools & capabilities (6)

When to use it

• Use it when you want an AI agent to answer business questions over governed Snowflake data via Cortex Analyst and semantic models
• Use it when building RAG over documents already indexed in Snowflake Cortex Search
• Use it when you need agents to create/describe/list Snowflake objects (databases, schemas, warehouses, stages) conversationally
• Use it when you want to run ad-hoc SQL from an assistant but gate dangerous statements via permission controls
• Use it when you want one server to expose Cortex AI plus SQL plus object admin instead of stitching together multiple tools

Quick setup

1. 1Create a `configuration.yaml` declaring your Cortex search/analyst/agent services and enabling object/query/semantic managers
2. 2Set Snowflake credentials as environment variables or CLI flags (account, user, and password/key-pair/OAuth)
3. 3Add the server to your MCP client config: command `uvx`, args `snowflake-labs-mcp --service-config-file config.yaml`
4. 4Restart the MCP client (Claude Desktop, Cursor, etc.) so it loads the server
5. 5Verify by asking the agent to list databases or run a simple Cortex Analyst question
6. 6For new projects, evaluate migrating to the official Snowflake Cortex MCP server referenced in the deprecation notice

Snowflake MCP Server (Snowflake Labs) FAQ