The Snowflake Tool enables Agents to connect to your Snowflake database and query data directly within chat conversations. The tool lets you retrieve information from your data warehouse without leaving Cassidy and makes it easy to access analytics, run queries, and get insights on demand.
A tool enables your Cassidy Agent to take a specific action in an external system, such as Snowflake. You can give someone access to different functions within your Snowflake Data Cloud. One tool might list all databases available in the account; another might list schemas within a database; and another might show tables, including metadata such as row counts and storage sizes. Each tool is a separate permission that you control individually.
When someone asks your Agent a question that requires Snowflake data, the Agent uses these tools to query Snowflake in real time and retrieve the information needed to respond. The data is not copied or stored in the Agent. It’s only accessed when needed to answer that specific question.
For best results with complex tools, use a powerful model. Your Agent works best with the following recommended models: Claude 4 Opus, Claude 4 Sonnet, GPT-5, or o3.
To set up and use the Snowflake Tool in your Agent, follow these steps:
Navigate to Agents: Click "Agents" in the left sidebar to access your Agent dashboard.
Select your Agent: Choose the Agent you want to add the Snowflake Tool to by clicking on it.
Access the Tools section: Click on “Edit” and navigate to the Tools section of the "Setup" tab to configure and manage tools available to your Agent.
Add Snowflake: Click "+ Add" and select "Snowflake" to enable this tool for your Agent.
Select your connection method: Choose between two authentication options:
Each user connects their own account (Recommended): Each team member links their own Snowflake account. The Agent only sees data that the user is authorized to access based on their Snowflake permissions. The setup works well for teams with different roles or restricted access in Snowflake.
Use a shared Snowflake connection: Create a service account with limited permissions for the workspace. Admins can control which data the Agent can access, independent of individual accounts. Only workspace admins can create or share the service connection, and it works well for uniform, centrally managed access.
Set up connection: Click “Connect” if you selected “Use a shared Snowflake connection.” You will be prompted to sign in, add your account ID, password, and username, and accept the access request.
Share the connection (optional): To share with teammates, click on “Change” next to the new connection, open the ellipses menu (…) and select “Share”.
Configure sharing permissions: Add people or groups, or set general access to “Team”. Assign specific access levels for each person or group.
Set tool permissions: Click “Done” and then select “Next” and choose a permission for each Snowflake action:
Always Allow: Agent runs the action automatically.
Needs Approval: Agent requests approval before running the action.
Disabled: Agent cannot run the action.
Common Snowflake tools include actions that let your Agent explore and interact with your Snowflake data platform. For example, these can include the tools for viewing metadata and querying data, such as Describe Table, Explore Columns, Show Databases, Show Schemas, Show Tables, and Run Read Query. They also include tools for modifying data, like Run Write Query. You decide the permission level for each tool separately. For example, you might set Show Databases and Run Read Query to “Always Allow”, but Run Write Query to “Disabled”. When you're done configuring permissions, click “Save”.
Start a chat: Go to Chat and select the Agent that has the Snowflake Tool enabled. If you choose the per-user option, users will be prompted in chat to connect their Snowflake account.
Your Agent can now query your Snowflake database to retrieve data, run analytics, and answer questions about your data warehouse directly in chat conversations.
Monitoring Snowflake actions
To review what the Agent did, open the “Analyzed Sources” dropdown and select any Snowflake step.
If an action is called and it doesn't have the permissions, it will not allow you to perform the action
If an action is called and needs approval, you will be able to click "Approve" or "Reject"
Data security
Permission-based access: The Agent only sees data that the connected account can access.
Auditability: Every query and action is logged in the Agent’s activity history.
Service account control: A shared service account lets admins restrict exactly which data the Agent can use.
Advanced: OAuth Scopes
This integration uses standard Snowflake OAuth scopes to secure access. If your organization needs custom scopes, contact [email protected].