Skip to main content
Query your Snowflake data warehouse with natural language, retrieve live data for reports, and quickly inspect table structures so you can build reliable, data-driven workflows in minutes.
Execute SQL queries and describe table schemas directly from your workflows, returning structured data that plugs into downstream nodes without custom parsing.

​
How to Use MCP Nodes

​
What is Snowflake MCP?

Snowflake MCP creates a customized node that understands how to work with your Snowflake warehouse. You can ask for table descriptions or run SQL using plain language, and the node translates that into the right tool call. Results come back as structured data that you can pass to other nodes in your workflow.

​
What Can It Do for You?

  • Run parameterized SQL and return structured rows for analytics, QA, and reporting
  • Inspect table schemas to confirm columns, types, and nullability before writing queries
  • Power repeatable workflows by returning structured data that connects to Sheets, Slack, and more
  • Rapidly prototype queries using natural language, then reuse the node across automations

​
Available Tools

ToolWhat It DoesExample Use
Describe TableDescribe the structure of a table in Snowflake”Describe schema name.table name and return structured data with columns: column_name, data_type, is_nullable”
Execute QueryExecute a SQL query on Snowflake”Run SELECT column list FROM schema name.table name WHERE filter condition and return structured data”

​
How to Use

1

Create Your Snowflake MCP Node

Go to your node library, search for Snowflake, and click “Create a node with AI”
2

Add Your Prompt

Drag the Snowflake MCP node to your canvas and add your prompt in the text box. Use variables like schema name, table name, and date range for reuse.
3

Test Your Node

Run the node to see the results. Snowflake MCP returns structured data, which you can pass to other nodes. If needed, iterate on the prompt.
4

Save and Reuse

Once your Snowflake MCP node is working, save it to your node library. You can now use this customized node in any workflow.

​
Example Prompts

Here are some prompts that work well with Snowflake MCP: Execute SQL Query: 
Given a SQL query, execute it and output the results
Query Data for Analytics: 
Run SELECT `order_id`, `order_date`, `total_amount`
FROM `schema name`.`orders`
WHERE `order_date` BETWEEN `start date` AND `end date`
AND `status` = `status value`
Return structured data with fields: order_id, order_date, total_amount
Check Table Schema Before Querying: 
Describe `schema name`.`customers` and return structured data with columns: column_name, data_type, is_nullable
Aggregation by Day: 
Run SELECT CAST(`date column` AS DATE) AS `day`, COUNT(*) AS `row_count`
FROM `schema name`.`table name`
WHERE `date column` BETWEEN `start date` AND `end date`
GROUP BY `day`
ORDER BY `day` ASC
Return structured data with fields: day, row_count
Preview Specific Columns: 
Run SELECT `column list`
FROM `schema name`.`table name`
WHERE `filter condition`
LIMIT `row count`
Return structured data with fields: `column list`
Start with a describe_table prompt to confirm column names and data types, then add a focused execute_query prompt that returns just the fields you need. Keep prompts parameterized with variables like schema name, table name, and date range for easy reuse in workflows.

​
Troubleshooting

If your Snowflake MCP node isn’t working as expected, try these best practices:

​
Keep Prompts Simple and Specific

  • Good: “Describe schema name.table name and return structured data with columns: column_name, data_type, is_nullable”
  • Bad: “Describe schema name.table name, then write a query to filter last quarter sales, join with customers, and export the results to Google Sheets”
While this prompt might work, it’s more efficient to break it into separate nodes. Snowflake MCP works best with focused, single-action prompts.

​
Match What Snowflake Can Do

  • Good: “Run SELECT customer_id, lifetime_value FROM schema name.customers WHERE lifetime_value > threshold and return structured data”
  • Bad: “Run a query, summarize the findings, and post the summary in Slack”
Snowflake MCP excels at database queries and schema inspection. For summaries, connect the output to an Ask AI node. For Slack messages, use Slack Message Sender in your workflow.

​
Break Complex Tasks Into Steps

Instead of trying to do everything in one prompt, which might cause timeouts or unnecessary complexity: 
Describe `schema name`.`orders`, join orders with `schema name`.`customers`, calculate weekly revenue for `date range`, format as a report, and write it to a spreadsheet
Break this into smaller, focused nodes that each handle one task:
1

Step 1: Describe Table

Describe schema name.orders and return structured data with columns: column_name, data_type, is_nullable
2

Step 2: Execute Query

Run SELECT DATE_TRUNC(‘week’, order_date) AS week, SUM(amount) AS weekly_revenue FROM schema name.orders WHERE order_date BETWEEN start date AND end date GROUP BY week ORDER BY week ASC Return structured data with fields: week, weekly_revenue
3

Step 3: Write to Sheets

Using the structured rows with fields week, weekly_revenue, write the data to Google Sheets using Google Sheets Writer with spreadsheet url and sheet name
In your workflow, connect these nodes sequentially. The table description validates columns used in the query step, and the structured rows from the query feed into the Sheets step for output.

​
Focus on Data Retrieval

Snowflake MCP is great at getting information from Snowflake. For analysis or content creation, connect it to other nodes. Example:
  • Good prompt: “Run SELECT date, SUM(revenue) AS total_revenue FROM schema name.orders WHERE date BETWEEN start date AND end date GROUP BY date and return structured data with fields: date, total_revenue”
  • Bad prompt: “Run a revenue query, analyze trends, and draft a summary paragraph”
Use Ask AI to analyze or summarize results. Keep Snowflake prompts focused on retrieving structured data that other nodes can use.

​
Troubleshooting Node Creation

If you’re seeing empty outputs in the node creation window (or if you’ve already created the node, hover over it and click “Edit”), use the chat interface to prompt the AI to add debug logs and verify the API response. For quick iteration, use the “Request changes” button in the node creation window and mention that you received empty outputs.
In the node creation window (or if you’ve already created the node, hover over it and click “Edit”), describe in detail what you expected versus what you received. Use the “Request changes” button in the node creation window to ask the AI to adjust filters, query logic, or output fields.
First click “Fix with Gummie”. If multiple attempts fail, simplify your prompt or contact support.
MCP node creation often benefits from a few tweaks. Use the chat interface in the node creation window to refine filters, output fields, or pagination. The AI will adjust the node based on your feedback.

​
Need More Help?