TL;DR: The best tool for data teams just got better. Introducing the Notebook Agent, bringing the best of agentic capabilities to where data teams are already working.
See it in action at our live virtual event on September 10th at 11 AM PT / 2 PM ET, and check us out on Product Hunt!
We’re starting to see agentic patterns transform software, most sharply in how software itself is made. AI code generation has become accurate and fast, and agent-augmented developer tools have become ubiquitous.
And yet in data, where writing code is a large part of the work, we haven’t seen the same tectonic shift. Instead, data practitioners are looking in envy as entire dev workflows are transformed — stuck copying and pasting, trying to keep up.
Why? Well, as it turns out, writing code was never the hardest part of the job!
Data analysis is not a verifiable task. It requires taste, intuition, and business judgment — things LLMs struggle with. A SQL query can execute successfully and still produce disastrously wrong results. The heart of the job lies in taking an analysis, reviewing its output, and turning it into institutional knowledge with real meaningful business impact.
That takes more than good code generation; you need purpose-built tools that can draw on governed context, integrate logic with analytics-specific tooling, and support collaborative iteration. And, just as critical as the analysis itself is transforming these insights into a shared story and artifacts that stakeholders can rely on for decision-making.
Agents, analytics, and you
Hex is already the preferred platform for thousands of data teams doing their best work. And now, it’s even better with a powerful analysis agent, built right into our notebook environment.
Start with any question, and the Notebook Agent leverages Claude Sonnet 4 and powerful tools in Hex to get you from question to first draft in minutes — abstracting away the most tedious parts of data work.
The Notebook Agent can help you do things like:
Plan and refine where you want to take your analysis
Search your data to quickly find the right table
Create and chain cells using SQL and Python for anything from a simple query to a complex multi-step analysis
Visualize your data with native pivot and explorable chart cells
Summarize and synthesize your findings in Markdown cells to share your insights
The Notebook Agent is available to Editors+ in public beta on all paid plans. AI usage is included for free during the early-beta period.
Built for data people
The Notebook Agent isn't just for one-off questions and first drafts — it's meant to help you iterate on complex data analysis. So we've invested in features that make it actually useful™️ for all of your production workflows.
@ mentioning cells & tables
When you know what you want, it’s nice to direct the agent to the right context. Easily @ mention tables, or tab to auto-add selected cells. Stay tuned for more control and context options to come!
Autosaving versions
Sometimes you want to go off-road and follow a train of inspiration while working on a project, but it doesn't always lead to a fruitful outcome.
That's why we built auto-saving for every new thread you started with the agent. Twenty messages and fifty cells later, realize there’s no good conclusion on an outlier data point? No problem! It's easy to restore a previous version.
Cell-by-cell revisions
Vibe analysis is cool, but precision matters when one wrong line could mean sending incorrect revenue numbers to your CEO. 😰
When the agent generates or modifies code cells, you'll see a clear diff view where you can edit suggestions inline and accept or reject changes at the cell level. Every line the agent produces is reviewable for a first-class editing experience.
Ready to dive in? Check out our Notebook Agent docs and best practices guide!
What’s next: more trusted context
Hex is increasingly the go-to place for folks outside the data team to ask and answer questions on their own — and the Notebook Agent will only accelerate that.
But data teams need a way to curate trusted context; relying on LLMs alone comes with too many gotchas! Last week, we launched semantic authoring, and next we'll be integrating agentic capabilities with semantic models, so anyone in your organization can ask questions in Hex using governed context straight from the data team.
The most powerful platform for agentic analytics
AI is only as useful as the context and tools behind it. The Notebook Agent lives in your Hex project with access to your code, existing files, and warehouse tables to understand your analysis. No copy-pasting, no managing local app deployments, and no cobbling tools together.
All the work you do with the agent becomes a durable, debuggable asset in Hex that your teammates can easily jump in and review. From there, it only takes a few clicks to turn your analysis into a polished, interactive data app for stakeholders to explore on their own.
The Notebook Agent is already supercharging data teams, and this is just the beginning. If you give it a spin, let us know your feedback — we’re eager to hear it!