Hex is the data workspace that won the workspace war. If you joined a data team after about 2022, there is a good chance the first link your manager sent you was a Hex notebook. It is, in 2026, the default tool for analysts who want to combine SQL, Python, and a shareable interactive interface in one place.

What It Actually Is

Hex is a browser-based notebook where each "cell" is one of: a SQL query against a connected warehouse, a Python (or R) snippet, a no-code chart builder, a markdown block, or an input widget like a dropdown or date picker. The killer property is that cells are reactive: when an upstream cell changes, every downstream cell that depends on it re-runs automatically, like a spreadsheet. The pile of broken Jupyter notebooks every data team has lying around exists because Jupyter is not reactive. Hex fixed this.

The other key property is the two-mode UI. Every Hex project has a "Logic view" (the notebook, with code visible) and an "App view" (a clean dashboard-like interface where the inputs become widgets and the outputs become a polished page). You build in Logic view, then publish the App view to a stakeholder URL. Stakeholders never see the code. Analysts never have to rebuild their notebook as a separate dashboard. This single feature is the reason Hex eats so much of the work that used to belong to BI tools.

The Origin Story

Hex was founded in 2019 by Barry McCardel (formerly at Palantir, where he ran data analytics work), Caitlin Colgrove, and Glen Takahashi. The founding insight came from watching Palantir analysts: the actual work of turning a question into a defensible answer was a messy mix of SQL, Python, charts, and prose, and there was no tool built to hold all of it together. Jupyter had the code part. Mode had the SQL part. Tableau had the charts part. Nobody had all three with the polish of a modern SaaS product.

Hex came out of stealth in 2020, raised aggressively on the strength of analyst love and a clean product, and by 2023 had become the obvious default for new data teams. Sequoia, a16z, Redpoint, and Amplify all participated in the rounds. The customer list grew to include the most visible AI-era companies (Notion, Anthropic, Reddit), which both demonstrated and amplified the product's quality.

Why Analysts Love It

Three things show up over and over in the love letters:

1. Reactivity makes notebooks trustworthy. In Jupyter or Mode, you can run cells in any order, end up in a state where the variables on screen don't match what the code says, and ship a wrong number to the CEO. In Hex, the dependency graph guarantees that what you see is what the code produced, in the right order, every time. This is the kind of feature you don't appreciate until you've been burned by its absence.

2. Hex Magic is good AI, not theater AI. Hex's AI assistant ("Magic") was integrated early — before most competitors had anything — and it was tightly coupled to the warehouse schema. It generates SQL that compiles, suggests Python that uses the actual columns in your dataframes, and writes documentation cells in the project's voice. Most "AI in the notebook" products in 2024-2025 felt bolted on. Magic felt native because it was.

3. Apps that ship. The fastest path from "an analyst is asked a question" to "a self-serve tool exists for that question" runs through Hex. Build the analysis as a notebook, add a few input widgets (date range, region dropdown), publish the App view, send the link. The app handles parameters, caches results, refreshes on a schedule, and looks good enough that PMs use it without complaint. This work used to require begging the BI team for a Looker model.

The Opinionated Take

Hex won the workspace category for the same reason Figma won design tools: the previous generation built single-player products and bolted collaboration on, while Hex built collaboration in from day one and made the single-player experience feel just as polished. Mode, the category creator, never escaped its identity as "a SQL-first reporting tool with a Python tab." Jupyter was an open-source local-first tool that nobody really managed at scale. Deepnote was technically excellent but never broke out of the data-science niche. Hex was built from the start as the canonical artifact — the analysis that lives at a URL, that anyone on the team can find, fork, and trust.

The other strategic choice that mattered: Hex did not try to become a BI tool. They could have. Instead they kept the workspace identity sharp — "this is for the people who write the code, and the people who consume their work" — and let Looker and Tableau own the executive dashboard layer. That focus is part of why the product feels coherent.

Where Hex Fits in the Stack

Hex sits between the warehouse and the human consumer. A typical Hex project queries Snowflake or BigQuery via a native connection, optionally references dbt models or a semantic layer, runs Python on the result, and publishes either as a notebook (for other analysts) or as an app (for stakeholders). Hex is not an ingestion tool, not a transformation tool, and deliberately not a governed metrics layer. It is the place analysis happens.

Pricing and Business Model

Hex is sold per-seat with separate pricing for "Editors" (analysts who write code) and "App users" (stakeholders who consume published apps). This pricing structure tells you something: Hex is betting that the consumer side of the workspace — the apps — is the eventual large surface. Editors are the wedge; apps are the expansion.