How to Get Started with Hex for Data Analysis

Hex is a cloud analytics workspace that runs SQL, Python, and no-code cells side by side in the same notebook, then lets you publish the result as an interactive app for stakeholders. If you have data in a warehouse or a spreadsheet and want to produce analysis and charts without setting up infrastructure, Hex is one of the shortest paths from raw data to a shared, filterable report.

Sign Up and Create a Workspace

Go to app.hex.tech and sign up with your work email. Hex offers a free tier suitable for individual analysts, with paid Team and Enterprise plans for larger groups. After signing in, you land in your workspace, which is the central hub for all your projects and data connections.

If you are joining a team that already uses Hex, have your admin send an invite rather than creating a new workspace. Hex does not support merging separate workspaces after the fact, so getting the invite step right saves headaches later.

For a new organization, two things matter before your first analysis: add a data connection, and optionally set up Context Studio so the AI Notebook Agent understands your business terminology. Both are covered in the steps below.

Connect Your Data

Click Settings in the left sidebar, then Data connections. Hex connects directly to Snowflake, BigQuery, Redshift, Databricks, PostgreSQL, MySQL, Athena, and more than 30 other warehouses. Each connection type prompts for the credentials appropriate to that service. Snowflake, for example, asks for account identifier, warehouse name, database, schema, role, username, and password.

If you do not have a warehouse, you can upload files directly into a project notebook. Hex supports CSV, Parquet, and Excel uploads. Uploaded files load into an in-memory DuckDB table inside the project, making them immediately queryable with SQL. For a quick proof of concept this is the fastest path.

Create one shared data connection at the workspace level rather than per-project connections. Shared connections mean any team member who creates a new project can start querying immediately, and credential rotation only happens in one place.

Create Your First Project

From the workspace home screen, click New project. A Hex project is a notebook that holds SQL cells, Python cells, text cells, chart cells, and input widgets, all sharing the same execution context. A variable defined in a SQL cell is accessible in a Python cell below it without any imports or data passing.

After creating a project, select your data connection from the dropdown in the top right. Then add a SQL cell by clicking the + button and selecting SQL. To confirm the connection is working, run a basic query:

SELECT * FROM your_table LIMIT 10;

If rows return, the connection is live and you are ready to start analyzing.

Hex evaluates cells top to bottom in sequence. If you add a Python cell that references the output of a SQL cell, run the SQL cell first. The Run All button at the top of the notebook handles this ordering automatically and is the safest way to refresh a full analysis.

Use the Notebook Agent

The Notebook Agent is the AI assistant for writing analysis logic. Activate it by pressing Cmd+G on Mac or Ctrl+G on Windows, or by clicking the Magic button at the top of any cell. The agent reads your live schema, the results of cells already run in the notebook, and any semantic context you have defined in Context Studio.

The quality of output depends heavily on prompt specificity. Two examples illustrating the difference:

Vague: "Show me my sales trends." The agent returns a generic query that may not match your actual schema or the metric you care about.

Specific: "Write a SQL query that calculates month-over-month revenue change as a percentage, grouped by product category, using the orders table with created_at as the date column and amount_usd as the revenue column, for the last 12 months." The agent writes a working query with correct column references and grouping logic on the first attempt.

When the result is close but not exact, describe the adjustment in natural language and the agent rewrites the cell. You do not need to edit SQL manually unless you prefer to.

On April 23, 2026, Hex released Context Suggestions, a feature that proactively surfaces relevant semantic definitions as you type a prompt. If your organization has defined "Monthly Recurring Revenue" in Context Studio, Hex surfaces that definition before you finish typing, reducing the risk of the agent calculating a metric differently than what your finance team considers correct. In organizations where multiple teams define similar metrics differently, this single feature eliminates a large category of analysis errors at the source.

Build Charts and Publish as an App

Once a SQL cell returns results, add a Chart cell below it. Hex auto-detects column types and suggests a chart type. You override the type in the right panel and configure axes, colors, and labels from dropdowns without writing code.

Hex supports bar, line, scatter, area, combo, and pivot layouts natively. For more control, render a chart in a Python cell using Matplotlib, Plotly, or Altair, and Hex displays the output inline in the notebook.

When the analysis is complete, click the App button in the top-right corner. This switches to the app view, which is a drag-and-drop layout for the cells the viewer sees. Toggle code cells to hidden, and promote charts, text, and input widgets to the front.

Input widgets let viewers change parameters without editing the notebook. A date range picker, a dropdown, or a text search box each maps to a SQL parameter. When a viewer changes the input, the query reruns and the chart updates. This is how you build a self-service report where a non-technical stakeholder can slice data themselves without a data analyst in the loop.

Click Publish to generate a URL. Team members access it in their browser. Public sharing is available if you turn off the login requirement in the sharing settings.

Set Up Context Studio for Better AI Answers

Context Studio is where you define the business logic the Notebook Agent draws on across all projects. Find it under Settings > Agent management > Context Studio.

Build a semantic model by selecting a data connection and defining two types of objects. Entities are key tables and their primary identifiers, for example Customer mapped to the customers table with customer_id as the key. Metrics are named calculations, for example Net Revenue defined as SUM(amount_usd) - SUM(refund_usd) from the orders table filtered to status = 'completed'.

Once defined, every new project using that data connection has the agent reference those definitions automatically. An analyst who prompts "show net revenue by customer segment" will get a query using your organization's specific revenue formula, not a generic SUM of whatever column looks like revenue.

This matters most in larger teams where inconsistent metric definitions are a recurring source of conflicting reports. Teams that set up Context Studio models spend less time correcting agent output and more time acting on the analysis.

If you want to analyze a CSV without a warehouse and skip this configuration entirely, VSLZ handles the same analysis from a file upload with a single plain-English prompt and no setup required.

What Hex Does Not Handle Well

Hex is built for internal analytics, not for embedding live dashboards in customer-facing products. The embedding feature exists, but the architecture is optimized for internal team use where controlling notebook access matters more than sub-second latency for anonymous visitors.

Hex is also not a data pipeline tool. You can schedule notebooks to rerun on a cadence, but a dedicated orchestration layer like dbt or Airflow is still necessary if you need structured extract-transform-load workflows and managed data dependencies. Hex sits downstream of that transformation layer.

The free plan limits projects and users, so teams that grow past a few analysts will hit the upgrade point before long.

Getting the Most Out of Hex

Start with one concrete analysis: pick a dataset you already understand, connect it, and run the analysis end to end. Write one real prompt to the Notebook Agent on a question you would normally answer manually. Publish the result as an app and share it with one stakeholder. That full loop takes under an hour and gives you a working mental model of where Hex fits in your workflow before you invest in deeper setup.