How to Get Started with Hex AI Analytics

Hex is a collaborative analytics notebook that runs entirely in the browser. You connect it to a data warehouse or upload a file, describe what you want in plain English, and Hex handles the query, chart, and summary. The Notebook Agent, introduced in late 2025 and powered by Claude Sonnet 4, can take a prompt like "show weekly revenue for the past 90 days, broken down by product category" and return working SQL, run it, and generate a chart in under a minute.

For analysts at companies that still circulate Excel files by email, Hex solves a specific problem: the gap between knowing what question you want answered and being able to produce the analysis. Traditional notebooks like Jupyter require a local environment, dependency management, and manual version control. Standard BI tools like Tableau or Looker require admins to build semantic models before analysts can query anything. Hex sits between the two. It feels like a notebook, delivers results like a BI tool, and requires almost no infrastructure to start.

As of April 2026, over 1,000 data teams use Hex. The platform shipped two AI updates in the past month: automatic user memory on April 14 (the agent retains your schema preferences across sessions) and Context Suggestions on April 23, which surfaces relevant tables and metrics as you type a prompt.

Step 1: Create an Account and Start a Project

Go to app.hex.tech and sign up with a work email. Hex offers a free tier with two projects and access to a built-in demo Snowflake dataset, which is enough to follow this guide end to end without connecting your own data.

Once inside, click the green + New Project button in the top-right corner. You land in the notebook view, the workspace where all analysis happens. The interface has three main areas: a left sidebar for data browsing and file uploads, a center canvas of cells, and a right panel for the Notebook Agent.

For a fast start, click the question mark icon in the bottom-left corner, choose Templates, and load the Dashboard template. This pre-built notebook includes sample SQL queries, charts, and input filters, and shows you what a finished project looks like before you build one from scratch.

Step 2: Connect Your Data

Hex supports direct warehouse connections to Snowflake, BigQuery, Redshift, Databricks, and PostgreSQL, as well as flat file uploads for CSV, Excel, and JSON, plus integrations with Google Drive, S3, and GCS.

For warehouse connections, an Admin in your workspace sets up credentials once under Settings > Data Connections, and all Editors inherit access. If you are setting up a personal workspace for the first time, the demo Snowflake connection called [Demo] Hex Public Data is already available in the Data browser. It includes sample tables covering orders, products, and customers, and you can run real queries against it immediately.

To upload a file instead, click the Files tab in the left sidebar, drag a CSV into the drop zone, and Hex registers it as a named dataframe. You can reference that dataframe in any SQL or Python cell that follows.

Step 3: Use the Notebook Agent

The Notebook Agent is the fastest path from a question to a working analysis. Open it by clicking the purple prompt bar at the top of the notebook canvas. Type a question in plain English, for example: "Using the orders table, calculate total revenue by month for 2025 and show a line chart."

The agent generates a SQL cell with the appropriate query, runs it against the connected data source, and adds a chart cell below it. The output is fully editable. Click into the SQL cell to modify the query, change the chart type, or ask a follow-up like "filter this to North America only."

A few prompting practices that produce better results, based on Hex's published documentation:

Start broad, then refine. Vague prompts produce structural scaffolding; specific prompts produce targeted edits. "Show me sales by region" gets you a working starting point. "Break out the Northeast and compare it to the same period last year" produces the revision you actually want.

Give the agent context about your schema. If your revenue column is called gmv_usd rather than revenue, say so in the prompt. Hex's automatic user memory, added April 14, 2026, retains these notes across sessions so you only need to teach the agent once per workspace.

Always validate the generated SQL. Check that GROUP BY logic is correct, that filters are not excluding unexpected rows, and that the output numbers pass a quick sanity check before you publish.

Step 4: Add Charts, Tables, and Text Cells

Once a query returns data, three cell types require no coding at all.

Chart cells configure bar, line, scatter, and area charts through a point-and-click menu. Map the x-axis to a date column, the y-axis to a numeric column, and the color grouping to a category. Hex renders an interactive chart that viewers can hover over to inspect values.

Table cells apply spreadsheet-style filters and aggregations to any dataframe. If your SQL query returns 50,000 rows, a table cell lets stakeholders filter by region or date without re-running the underlying query.

Markdown cells add narrative context between data cells. Use them to label sections, explain methodology, or call out a key finding above a chart. In a published app, these cells become the structure that turns a notebook into a readable report.

Step 5: Publish as a Shareable App

When the analysis is complete, click App Builder at the top of the screen. Hex lets you drag and drop cells from the notebook into an app layout, hide the underlying code, and add interactive input controls. A dropdown filter cell, for example, lets a stakeholder narrow the report by product category without editing any SQL.

Click Publish in the top-right corner. Hex generates a shareable URL. Recipients with Viewer access can open the app in a browser, interact with filters, and export results to CSV, but cannot see or modify the notebook. This separation between development and consumption makes Hex practical for teams where analysts build and non-technical stakeholders consume.

Hex's free tier includes two published apps. The Team plan is priced at $24 per user per month as of early 2026 and removes that cap, adding full Notebook Agent access and scheduled runs.

Scheduling Automated Refreshes

If you want the analysis to update without manual intervention, the Scheduled Runs feature on paid plans can re-execute the entire notebook daily or hourly and push updated results to the published app URL automatically. This turns a one-time analysis into a living report that stakeholders can bookmark and check each morning.

Practical Notes

The Notebook Agent requires Editor role or above on a paid plan; Viewer accounts cannot run it. Hex stores query results in session memory, not on disk, so opening a saved project does not restore computed state. Run all cells from the top before sharing an app to ensure the data is current.

For teams that want to skip warehouse configuration entirely and go straight to querying their data from a file upload, VSLZ handles this through a single upload step with no connection setup required.