Skip to main content
The Snowflake connector enables natural language querying of your Snowflake data warehouse through Cortex Analyst. Users can ask questions in plain English and get results without writing SQL.

​
What you can do

  • Natural language queries - Ask questions like “What were our top 10 customers by revenue last quarter?” and get results
  • Direct SQL queries - Execute read-only SQL queries against your Snowflake warehouse
  • Semantic models - Register YAML-based semantic models or Snowflake semantic views to define your data structure

​
Prerequisites

Before connecting Snowflake to Grapevine, you need:

​
Setup

Create a dedicated read-only role for Grapevine queries. Run the following SQL in Snowflake: 
-- Create a role for Grapevine
CREATE ROLE IF NOT EXISTS GRAPEVINE_ROLE;

-- Grant warehouse usage
GRANT USAGE ON WAREHOUSE <your_warehouse> TO ROLE GRAPEVINE_ROLE;

-- Grant database and schema access
GRANT USAGE ON DATABASE <your_database> TO ROLE GRAPEVINE_ROLE;
GRANT USAGE ON SCHEMA <your_database>.<your_schema> TO ROLE GRAPEVINE_ROLE;

-- Grant read access to tables
GRANT SELECT ON ALL TABLES IN SCHEMA <your_database>.<your_schema> TO ROLE GRAPEVINE_ROLE;
GRANT SELECT ON FUTURE TABLES IN SCHEMA <your_database>.<your_schema> TO ROLE GRAPEVINE_ROLE;

-- Grant access to stages (if using YAML semantic models)
GRANT READ ON STAGE <your_database>.<your_schema>.<your_stage> TO ROLE GRAPEVINE_ROLE;

-- Grant Cortex access
GRANT DATABASE ROLE SNOWFLAKE.CORTEX_USER TO ROLE GRAPEVINE_ROLE;

-- Grant the role to your user
GRANT ROLE GRAPEVINE_ROLE TO USER <your_username>;
Replace placeholders (<your_warehouse>, <your_database>, etc.) with your actual Snowflake object names.

​
2. Create an OAuth security integration

Create an OAuth integration in Snowflake for Grapevine authentication: 
CREATE OR REPLACE SECURITY INTEGRATION GRAPEVINE_OAUTH
  TYPE = OAUTH
  ENABLED = TRUE
  OAUTH_CLIENT = CUSTOM
  OAUTH_CLIENT_TYPE = 'CONFIDENTIAL'
  OAUTH_REDIRECT_URI = 'https://auth.getgrapevine.ai/callback/snowflake'
  OAUTH_ISSUE_REFRESH_TOKENS = TRUE
  OAUTH_REFRESH_TOKEN_VALIDITY = 7776000  -- 90 days
  PRE_AUTHORIZED_ROLES_LIST = ('GRAPEVINE_ROLE');

​
3. Get your OAuth credentials

Retrieve the client ID and secret: 
SELECT SYSTEM$SHOW_OAUTH_CLIENT_SECRETS('GRAPEVINE_OAUTH');

​
4. Connect to Grapevine

1

Open Snowflake settings

In Grapevine, go to Integrations and click on Snowflake.
2

Enter your credentials

Enter your Snowflake Account Identifier (e.g., myorg-account123 from your Snowflake URL).For the OAuth credentials, you have two options:
  • Quick setup - Paste the JSON output from SYSTEM$SHOW_OAUTH_CLIENT_SECRETS to auto-fill the Client ID and Secret
  • Manual entry - Enter the OAuth Client ID and OAuth Client Secret individually
3

Authorize

Click Connect with OAuth and authorize Grapevine in the Snowflake login prompt.

​
5. Register a semantic model

After connecting, register at least one semantic model to enable natural language queries.
If you have a YAML semantic model file stored in a Snowflake stage:
  1. Click Add Semantic Model
  2. Select Semantic Model (YAML file in stage)
  3. Enter a Name for the model
  4. Select a Stage Path, then edit the path to include your YAML file (e.g., @MY_STAGE/models/sales.yaml)
  5. Add a Description (optional)
  6. Select a Warehouse for query execution
  7. Click Add Model
Adding a description helps the agent understand what data is available in each semantic model, improving the quality of natural language query responses.

​
Using the connector

Once configured, users can query Snowflake data through the Grapevine MCP server.

​
Natural language queries

Ask questions in plain English. Cortex Analyst translates them to SQL and returns results: 
What were our top 10 products by revenue last month?

Show me the average order value by customer segment for Q4 2024

​
Direct SQL queries

You can also execute read-only SQL directly: 
SELECT product_name, SUM(revenue) as total_revenue
FROM sales
WHERE order_date >= '2024-01-01'
GROUP BY product_name
ORDER BY total_revenue DESC
LIMIT 10;

​
Permissions

  • Queries execute using the OAuth user’s Snowflake permissions
  • Snowflake role-based access control (RBAC) is respected
  • Users only see data they have access to in Snowflake
  • All queries are read-only (SELECT only)

​
Sync behavior

Unlike other connectors, Snowflake does not index data. Queries are executed in real-time against your Snowflake warehouse.
  • No data stored - Query results are not persisted in Grapevine
  • Live queries - Each query runs against current Snowflake data
  • Token refresh - OAuth tokens are automatically refreshed

​
Limitations

  • Natural language queries require a configured semantic model
  • SQL queries are read-only (no INSERT, UPDATE, DELETE)
  • Query results are limited to 1,000 rows by default
  • Large result sets may be truncated
  • Cortex Analyst availability varies by Snowflake region