Skip to main content

Connect Snowflake Cortex AI to the Knak MCP Server

Connect your Snowflake Cortex AI agent to the Knak Model Context Protocol (MCP) server. This lets your Snowflake agent access Knak through MCP.

Before You Begin

Make sure you have:

  • Access to a Snowflake account with permission to create API integrations.

  • Access to create or edit Snowflake Agents.

  • Access to your Knak account.

  • The Knak MCP server URL:

    • https://enterprise.knak.io/mcp/public

This feature is currently in Alpha Testing. Contact grace.lee@knak.com to request access.

Do not proceed with MCP setup until access has been granted. The MCP connection will fail if this step is not completed first.

Create the Knak MCP Server

Create the API integration first. Then create the external MCP server.

Note: You cannot create a custom MCP connector entirely from the Snowflake user interface (UI). You must first create the connector by using Snowflake Structured Query Language (SQL).

  1. Open a Snowflake SQL worksheet.

    +Add New menu expanded with SQL file option visible.

  2. Replace YOUR_SCHEMA with your schema name and replace YOUR_ROLE with a role.

  3. Run the following SQL:

    USE SCHEMA YOUR_SCHEMA;

    SET KNAK_MCP_URL = 'https://enterprise.knak.io/mcp/public';

    CREATE API INTEGRATION KNAK_MCP_API_INTEGRATION
      API_PROVIDER = external_mcp
      API_ALLOWED_PREFIXES = ($KNAK_MCP_URL)
      API_USER_AUTHENTICATION = (
        TYPE = OAUTH_DYNAMIC_CLIENT
        OAUTH_RESOURCE_URL=$KNAK_MCP_URL
      )
      ENABLED = TRUE;

    GRANT USAGE ON INTEGRATION KNAK_MCP_API_INTEGRATION TO ROLE YOUR_ROLE;

    CREATE EXTERNAL MCP SERVER KNAK_MCP_SERVER
      WITH DISPLAY_NAME = 'Knak MCP Server'
      URL = $KNAK_MCP_URL
      API_INTEGRATION = KNAK_MCP_API_INTEGRATION;

    GRANT USAGE ON EXTERNAL MCP SERVER KNAK_MCP_SERVER TO ROLE YOUR_ROLE;

Verify the Result

After the SQL finishes running, Snowflake displays a confirmation message similar to:

MCP server KNAK_MCP_SERVER successfully created.

This message confirms that the Knak MCP Server was created successfully and is ready to add to a Snowflake Agent.

Snowflake SQL worksheet showing the message "MCP server KNAK_MCP_SERVER successfully created."

Add the MCP Server to a Snowflake Agent

After you create the MCP server, add it to your Snowflake Agent.

  1. Open the Snowflake application.

  2. Select AI & ML.

  3. Select Agents.

  4. Create a new agent or open an existing agent.

  5. Select the MCP Connectors tab.

  6. Find Knak MCP Server.

  7. Click Add to Agent.

    The MCP Connectors tab with Knak MCP Server highlighted.

Authorize the Connector in Snowflake CoWork

After adding the connector, authorize it through Knak.

  1. In Snowflake, select AI & ML.

  2. Select Snowflake CoWork.

    Snowflake cowork navigation visible.

  3. Sign in if prompted.

  4. In the prompt input field, click the + button.

    Action menu expanded in the input field section of Snowflake Cowork.

  5. Select Connectors.

  6. Click Connect beside Knak MCP Server.

  7. Complete the authorization flow in Knak.

    Knak authorization screen

  8. Return to Snowflake CoWork and verify your connection.

    Connected status after authorization

After authorization completes, your Snowflake Agent can use the Knak MCP server.

Verify the Result

After the SQL completes, Snowflake displays a success message for each statement.

If all statements run successfully, the Knak MCP Server is ready to use.

Need more help? Contact support via live chat using the chat bubble in the bottom right corner or email support@knak.com.

Related Articles
Did this answer your question?