Snowflake Agent

A simple Snowflake subject-matter expert agent that answers business questions, runs SQL against Snowflake, explains results in plain English, and helps users explore operational data without writing queries manually.

​

Tutorial Summary

• Goal: Build a Snowflake SME agent that can query support-ticket data and explain insights clearly.
• Estimated Time: 10-15 minutes
• What you’ll build: A Snowflake-connected agent that understands your default database context, answers operational questions, and summarizes results from the SUPPORT_TICKETS table.

​

Key Features

• Native Snowflake Integration
• Natural-language Snowflake Q&A
• Business-friendly summaries with tables and takeaways
• SLA-risk ticket analysis, Resolution-time and team workload reporting

​

Prerequisites

• xpander.ai account
• Snowflake Warehouse
• Snowflake Programmatic Access Token
• Database and table created in Snowflake

​

Step-by-Step Implementation

​

Step 1 - Create the Snowflake Agent

In xpander.ai, create a new agent and name it: Snowflake Agent This agent acts as a lightweight Snowflake SME for answering data questions over support-ticket data.

​

Step 2 - Connect Snowflake

Add the Snowflake connector and configure it using API Key authentication. Use:

• Auth: API Key
• Scheme: Bearer
• API Key: Your Snowflake Programmatic Access Token
• Snowflake URL: https://<your-account-identifier>.snowflakecomputing.com

​

Step 3 - Add SME Instructions

Use these instructions to make the agent answer like a business-focused Snowflake SME:

You are a Snowflake SME agent. Your job is to help users answer business questions using Snowflake data. Use the default Snowflake context:

• Role: <your-snowflake-role>
• Warehouse: <your-snowflake-warehouse>
• Database: <your-database-name>
• Schema: <your-schema-name>

Primary table: <your-database-name>.<your-schema-name>

​

Example Questions

SLA-risk ticket analysis: The agent compares average resolution time across products: The agent summarizes product-level ticket health and support trends:

​

Troubleshooting

1. Wrong Snowflake URL: Use the full Snowflake account URL, not just the short account name.
2. Invalid or expired token: Generate a new Programmatic Access Token and update the connector.
3. Missing network policy: Grant temporary access for the token or configure a valid Snowflake network policy.
4. Agent asks for database details: Add the default role, warehouse, database and schema to the agent instructions.
5. Permission error: Make sure the connected Snowflake user has access to the warehouse, database, schema, and tables.
6. Query does not run: Confirm the warehouse is active and the selected role has permission to use it.

​

Conclusion

In this tutorial, you built a Snowflake SME Agent that connects to Snowflake, understands a default database context, and answers support-ticket questions in plain English. The agent can identify SLA-risk tickets, calculate resolution-time metrics, summarize team workload, and explain results with business-friendly takeaways. Although this example uses a small demo table, the same pattern can be extended to larger Snowflake environments by adding more schemas, richer business rules, and specialized instructions for different data domains.