Cortex Code CLI

This topic helps you get started with Cortex Code CLI, including installation, connection setup, and validation.

Before you begin, ensure you have a Snowflake account with access to the required Cortex models. See Prerequisites for full details.

Note

If you do not have a Snowflake account, you can sign up for a free Cortex Code CLI trial (https://signup.snowflake.com/cortex-code).

Install Cortex Code CLI

To install Cortex Code CLI, issue the following command in your terminal:

curl -LsS https://ai.snowflake.com/static/cc-scripts/install.sh | sh
Copy

This command downloads and runs the installation script, which installs the latest version of Cortex Code CLI. The cortex executable is installed in ~/.local/bin by default.

Connect to Snowflake

After installing the Cortex Code CLI, issue the cortex command. A setup wizard guides you through the initial configuration steps, including choosing or setting up a connection to Snowflake.

The first prompt asks you to choose a connection from the existing connections in the ~/.snowflake/connections.toml file or to create a new connection.

  • To use an existing connection, choose the connection from the list using the up and down arrow keys, then press Enter.

  • To create a new connection, choose More options* by pressing the down arrow key until it is highlighted, then press Enter. Follow the prompts to enter your Snowflake account details.

Note

The connections.toml is also used by the Snowflake CLI (snow command). If you have already set up a connection for use with the Snowflake CLI, you can use that connection with the Cortex Code CLI.

Start using Cortex Code

Once connected, try your first request:

What can I do with Cortex Code?
Copy

Type natural-language requests (such as “find tables with PII tags” or “generate a Streamlit app for SALES_MART.REVENUE”) and Cortex Code attempts to fulfill the request by orchestrating Snowflake-native skills and any MCP tools you have configured. For more information on configuring MCP tools, see Model Context Protocol (MCP).

As it works on your request, Cortex Code CLI displays its reasoning steps and actions in the terminal. From time to time, it may ask you for information that it needs. If you’re in plan mode, it will ask you to confirm each action.

Example requests

Discover your catalog

What databases do I have access to?
List every table tagged PII = TRUE in ANALYTICS_DB
Show the lineage from RAW_DB.ORDERS to downstream dashboards
Copy

Generate and run SQL commands

Write a query for top 10 customers by revenue
Add a 7-day moving average and show me the results
Explain why this query is slow and optimize it
Copy

Build applications

Build a Streamlit dashboard on SALES_MART.REVENUE with filters for date and region
Create a dbt project to transform raw sales data
Copy

Work with Cortex Analyst

Use the @models/revenue.yaml semantic model to answer "What was revenue last month?"
Debug my semantic model at @models/revenue.yaml
Copy

Prerequisites

To use Cortex Code CLI, you need the following:

  • A Snowflake user account with the necessary permissions to access the data you intend to use with Cortex Code CLI and to perform operations on them. This user must also have the SNOWFLAKE.CORTEX_USER database role.

  • Network access to your Snowflake server.

  • Snowflake CLI installed on your workstation.

  • One of the following supported platforms:

    • macOS on Apple Silicon or Intel

    • Linux on Intel

    • Windows Subsystem for Linux (WSL) on Intel

    Note

    Snowflake may add support for other platforms from time to time. Please let your Snowflake representative know if you have a specific platform requirement.

  • Local terminal access to the bash, zsh, or fish shell on your platform.

For additional configuration options, troubleshooting, and advanced setup, see Cortex Code CLI reference.

Supported models and regions

Cortex Code CLI supports the following models, as long as your account has access to them (see Control model access):

  • auto

  • Claude Sonnet 4.5 (claude-sonnet-4-5)

  • Claude Sonnet 4.0 (claude-4-sonnet)

  • Claude Opus 4.5 (claude-opus-4-5)

The default model is auto, which automatically uses the best available model subject to region restrictions and permissions. To choose a different model, use the /model command inside a Cortex Code CLI session.

If a model you want to use is not available in your region, you can use Cortex cross-region inference to access the model in another region where it is available. For more information about configuring cross-region inference, see Cross-region inference.

Important

Cross-region inference is required when the selected model is not available in your region. If inference fails with a model availability error, configure cross-region inference:

  • AWS US - Claude Sonnet 4+ offers the highest quality. Set up Cortex Cross-region inference for AWS_US to access Claude Sonnet 4.x models.

  • AWS EU - Set up Cortex Cross-region inference for AWS_EU to access Claude models.

  • AWS APJ - Set up Cortex Cross-region inference for AWS_APJ to access Claude models.

  • Any region - Set up Cortex Cross-region inference for ANY_REGION to access all models.

To enable cross-region inference, an ACCOUNTADMIN must run:

ALTER ACCOUNT SET CORTEX_ENABLED_CROSS_REGION = 'AWS_US';
Copy

Replace AWS_US with the appropriate region identifier (AWS_US, AWS_EU, AWS_APJ, ANY_REGION).

Your organization can restrict model access. You may not have access to all models. See Control model access for details.

Language: English