Modern organizations face the challenge of managing two types of data:

The ability to analyze and derive insights from both types of data is crucial for understanding customer needs, improving processes, and driving business growth.

In this quickstart, you'll learn how to build an Intelligent Sales Assistant that leverages Snowflake's capabilities for analyzing both sales conversations and metrics. Using Cortex Code CLI — Snowflake's AI-powered command-line assistant — we'll create, configure, and test the agent entirely from the terminal.

Snowflake Cortex AI allows you to turn your conversations, documents and images into intelligent insights with AI next to your data. You can access industry-leading LLMs at scale directly in SQL or via APIs, analyze multimodal data and build agents — all within Snowflake's secure perimeter. The platform leverages powerful capabilities:

Use AI to accurately convert natural language to SQL; Simple to use, cost-efficient to scale.

Find information by asking questions within a given set of documents with fully managed text embedding, hybrid search (semantic + keyword) and retrieval.

Orchestrate across both structured and unstructured data to retrieve and synthesize high-quality data insights. The Cortex Agents is a stateless REST API endpoint that:

Cortex Code is Snowflake's AI-powered command-line interface for building, testing, and managing Snowflake resources conversationally. It includes built-in skills — structured workflows that guide you through complex tasks like agent creation, evaluation, and optimization — all from your terminal.

These capabilities work together to:

1. Search through sales conversations for relevant context
2. Go from Text to SQL to answer analytical questions
3. Combine structured and unstructured data analysis
4. Provide natural language interactions with your data

A full-stack application that enables users to:

If you don't already have Cortex Code CLI installed, run the following in your terminal:

macOS and Linux (including WSL):

Windows (PowerShell):

After installation, verify it works and connect to your Snowflake account:

On first launch, Cortex Code will guide you through connecting to your Snowflake account. Ensure your connection uses a role with privileges to create databases, schemas, tables, stages, Cortex Search services, and agents.

We'll use Cortex Code to execute the SQL that creates our sales intelligence database. Clone the companion repository first:

Now launch Cortex Code from the repo directory:

Step 1. Ask Cortex Code to run the setup SQL. You can paste this prompt:

Cortex Code will execute each statement from setup.sql, which:

Step 2. Upload the semantic model. The Agent needs rules to interpret your structured data. This semantic model defines key business terms to ensure the Agent generates highly accurate SQL. Ask Cortex Code:

Cortex Code will run:

Now we'll use Cortex Code's built-in cortex-agent skill to create the Sales Intelligence Agent. This skill provides a guided, interactive workflow for agent creation.

Step 1. In Cortex Code, ask it to create the agent:

Cortex Code will load the cortex-agent skill and walk you through the creation workflow interactively.

Step 2. When prompted for administrative setup, provide:

Step 3. When prompted for the agent's purpose and requirements, explain:

Step 4. When prompted for tools, select the existing tools in the SALES_INTELLIGENCE.DATA schema:

Step 5. When prompted for response instructions, provide:

Step 6. When prompted for orchestration instructions, provide:

Step 7. Cortex Code will create the agent via the REST API and verify its configuration. You can confirm the agent was created by asking:

Step 8. Grant access to the role. Ask Cortex Code:

Now that your Agent is fully configured, let's test it directly from Cortex Code. Ask Cortex Code to chat with the agent:

Cortex Code will load the chat-with-agent capability. Try these questions to test both structured and unstructured data:

Test 1 — Structured data (Cortex Analyst):

This question requires the Agent to run a SQL query against the sales metrics table using Cortex Analyst.

Test 2 — Unstructured data (Cortex Search):

This question requires the Agent to look up contextual information within the long text transcripts in the sales conversations table using Cortex Search.

Test 3 — Combined:

This tests the agent's ability to orchestrate across both tools.

In addition to testing in Cortex Code, you can integrate your Agent into custom applications. You can interact with the Agent by calling the Snowflake REST API at /api/v2/databases/{DATABASE}/schemas/{SCHEMA}/agents/{AGENT}:run.

We've created a simple Streamlit application that interacts with the REST API here. You can use it as a building block to build your own applications.

Since this API call is made outside of Snowsight, we need a secure key to authenticate your identity and grant the application permission to run the Agent. This key is a Programmatic Access Token (PAT).

Let's run the Streamlit app and call the API locally:

Step 1. Create a Programmatic Access Token. You can do this via Cortex Code:

Or manually in Snowsight: Click on your profile (bottom left corner) → Settings → Authentication → Under Programmatic access tokens, click Generate new token → Select Single Role and select sales_intelligence_role. Copy and save the token.

Step 2. Find your account URL. Ask Cortex Code:

Or in Snowsight: Click on your profile (bottom left corner) → Account → View account details.

Step 3. Set up and run the Streamlit app. In the cloned repository directory, run:

Replace <PAT> with the token from Step 1 and <ACCOUNT_URL> with your account URL from Step 2.

When you're done using the application, press Ctrl+C in the terminal to shut down the Streamlit server.

Congratulations! You've successfully built an Intelligent Sales Assistant using Snowflake Cortex Agents — entirely from the command line with Cortex Code CLI. This application demonstrates the power of combining structured and unstructured data analysis through: