How to Get Started with Hex for Team Analytics

Getting Started

Hex is a cloud-based analytics workspace that combines SQL, Python, and no-code cells in a single collaborative notebook environment. Teams using Hex connect directly to their data warehouse, build analyses with mixed SQL and Python, and publish the results as interactive apps for non-technical stakeholders. To get started, sign up at hex.tech, connect a data source, and open a new project. The Notebook Agent can generate your first SQL query and chart from a plain English description within minutes.

Why Teams Switch to Hex

Most analyst teams run into the same friction with Jupyter notebooks: files run locally, sharing means emailing .ipynb files, and business stakeholders cannot interact with the output. Hex solves this by hosting notebooks in the cloud with live multiplayer editing, permission controls, and a one-click app view that turns any analysis into an interactive dashboard without separate dashboarding software.

The platform has three main areas: the notebook (called Logic View), the App builder, and the Agent Management section for configuring AI context. Each is accessible from tabs at the top of any project.

Step 1: Create Your Workspace

Go to app.hex.tech/signup and create an account. Hex offers a free trial. After signup, the workspace home screen loads with a sample project you can open immediately to see the interface.

To invite teammates, go to Settings, then Users, and send email invitations. Hex distinguishes between Viewers, who can read and interact with published apps, and Editors, who can build notebooks and use the Notebook Agent. Editors consume credits on paid plans. For a small team starting out, two or three Editor seats with Viewer access for everyone else is a reasonable starting configuration.

Step 2: Connect a Data Source

Click Add a data connection from the sidebar. Hex has native connectors for Snowflake, BigQuery, Redshift, PostgreSQL, MySQL, and Databricks SQL warehouses. For Snowflake, you enter your account identifier (formatted like xy12345.us-east-1), warehouse name, database, schema, and credentials. Hex tests the connection and populates the Data Browser with your schema tree once connected.

File uploads are the fastest path to your first analysis. CSV, Excel, and Parquet files up to 500MB can be uploaded directly inside a project by going to New Project, then Upload file. Hex infers column types automatically and exposes the file as a DataFrame named df.

Step 3: Build Your First Notebook

A Hex project is a sequence of cells. The four cell types you will use most often are SQL cells, Python cells, chart cells, and input cells.

SQL cells run standard SQL against your connected warehouse. Each SQL cell output becomes a named DataFrame, accessible by name in later Python cells. A query like SELECT date, revenue FROM orders WHERE date > '2024-01-01' creates a df_orders DataFrame automatically, which you can then manipulate in Python.

Python cells support standard Python with pandas, numpy, matplotlib, and most common data libraries pre-installed. No environment setup is needed. You reference DataFrames from SQL cells by name.

Chart cells are a no-code chart builder. Select a DataFrame, choose a chart type (bar, line, scatter, table, and several others), and configure axes with dropdowns. No code is required to build a working chart.

Input cells add interactivity: date pickers, dropdowns, and sliders. Connect an input cell to a SQL cell parameter and the query re-executes automatically when a viewer adjusts the input. A date range picker that filters a revenue query WHERE clause is a common first use.

A typical workflow chains these together: SQL cell pulls raw data, a Python cell cleans or aggregates it, a chart cell visualizes the result, and an input cell lets stakeholders filter by region, date, or product category.

Step 4: Use the Notebook Agent

The Notebook Agent is Hex's built-in AI assistant that operates directly inside the notebook. It launched in August 2025 as part of Hex's Fall 2025 release and received substantial updates in December 2025. To open the agent, press the lightning bolt icon in the toolbar or use Command-K on Mac.

The agent handles four main tasks. It generates complete SQL queries from plain English descriptions, for example "show me monthly revenue by product category for the last 12 months." It writes Python code to clean or transform DataFrames. It creates and styles charts automatically. And it can generate an entire multi-cell analysis from a single prompt when given enough context about the data.

For best results, configure Context Studio before prompting heavily. In Agent Management, Context Studio lets you write a brief description of your tables, define key metrics (for example, "revenue means net revenue after refunds, not gross"), and specify valid join keys. Teams that configure semantic models before scaling see substantially fewer correction rounds on agent-generated queries. Hex's internal benchmarking shows that semantic model context reduces the number of query corrections needed by roughly 40 percent compared to cold-start prompts with no context provided.

The agent is available to Editors on all paid plans.

Step 5: Publish as an Interactive App

Once your analysis is working, switch to the App Builder tab. Hex provides a drag-and-drop canvas where you arrange charts, tables, and input widgets. Code cells that stakeholders do not need to see can be hidden.

Click Publish, then Share, to generate a link. Viewers open the link, adjust input parameters, and see charts update in real time without ever seeing the underlying SQL or Python. Published apps can be embedded in Notion, Confluence, or any tool that accepts iframes.

Scheduled runs keep apps refreshed automatically. In Run settings, choose a frequency (hourly, daily, or weekly) and Hex re-executes the notebook and pushes updated results to everyone holding the link.

Step 6: Collaborate in Real Time

Hex supports live multiplayer editing. Two analysts can work in the same notebook simultaneously, with each collaborator's cursor labeled by name. Changes appear instantly.

For async review, inline comments let anyone highlight a cell, chart, or output and start a thread. The team resolves comments before publishing. This is especially useful when a business stakeholder wants to question a metric definition without rewriting the analysis.

The Reviews feature gates publishing behind an approval workflow. An analyst drafts the notebook, a senior reviewer approves it, and only approved versions go live. For recurring reports where accuracy matters, this prevents early drafts from reaching stakeholders.

When Hex Is Not the Right Fit

Hex works well when a team includes both SQL or Python users who build analyses and non-technical stakeholders who need to interact with the output. It is less suited for heavy data engineering tasks like running dbt models or orchestrating pipelines, where dedicated tools are faster. It is also not designed for real-time streaming dashboards that update at sub-second frequency.

For teams that want to skip the notebook setup entirely and simply ask questions of their data in plain English, VSLZ lets you upload a CSV or connect a data source and get analysis, statistical breakdowns, and charts from a single prompt with no configuration required.

Practical Tips for Getting the Most Out of Hex

Use query mode instead of dataframe mode for large datasets. In query mode, data stays in your warehouse and only the result set is returned, which avoids memory issues on tables with millions of rows and keeps execution fast.

Build semantic models before scaling to a full team. Defining metric calculations and valid joins in Context Studio takes about five minutes per table set and pays back every time the Notebook Agent generates the correct answer on the first attempt rather than the third.

Use components for shared analyses. If three projects all need the same revenue calculation, build it as a Hex component and import it into each project. When the definition changes, one update propagates everywhere.

Keep published apps focused. App Builder supports many charts, but published apps load faster and get more consistent engagement when they present the three or four charts that answer the core business question. Save the full exploration notebook for internal use.

Starting Your First Project

Getting from signup to a working published analysis in Hex takes approximately 15 to 20 minutes: create an account, connect a data source, build a notebook with SQL and chart cells, and publish it as an interactive app. The Notebook Agent reduces the time to a working SQL query significantly, especially once semantic context is in place. For analyst-to-stakeholder workflows where the output needs to be interactive and shareable, Hex eliminates the usual friction of Jupyter files, scheduled email exports, and static PDFs.