"Claude Dashboards: The Jupyter Notebooks of Agent Age"

Claude Dashboards were recently announced, and it goes to show that Dashboard are indeed not dead - when Claude has now built their own services.

The integrations is done into these Cloud Data Platform:

Claude Dashboards connects to a data platform like BigQuery, Databricks or Snowflake, or a CRM tool like Salesforce, and is in beta on paid plans.

This is following the trend of dbt Charts, MotherDuck Dives and other BI-as-Code first tools such as Rill Data or Evidence.

# What Claude Dashboards Can’t Do: Disadvantages

It’s like “live data artifacts” but without you being in control.

Christophe Blefari said it best:

You can decide to use Claude Dashboards and build everything on top or you can use the open-source version of it that we are developing. With nao you can do everything you can do with Claude Dashboards and:

• managing and self-improving you context layer in one UI versioned with Git
• using other LLM providers (even the open-source ones)
• having a central governance
• Use RLS on your data
• self-host it in your own infrastructure
• improve the product or even extend it for your own requirements

The nao stands for any BI-as-Code dashboarding tool, you need access to the artifacts to integrate it into the data and open standards. Tightly integrated into Claude will be OK for quick exploration or examples, but not for long-term managed dashboards

# Claude Dashboards Are the Jupyter Notebooks of the Agent Age

Jupyter Notebooks are great for creating a quick exploration or sharing a deep analysis. Still, they are not great for integrating into your data stack or your data pipelines. The same is now with Claude Dashboards.

# BI-as-Code is Still the Future

Anything code-first is what will win in long-term. It has shown once again; it first started with Markdown with Agentic Skills, Specs that we define with superpowers and others Spec Driven Development (SDD). T

Then Claude went to creating HTML dashboards within Claude, and then today we have full-blown charts. But without its code, or being open source, you are locked-in once again, and you can’t develop or, even more ** important, maintain it over time**. So it just gets stale after a very short time.

A decade ago we discussed no-code vs. code, now it’s in-built in frontier models/apps vs. open source or code-first. Once again, IMO, it’s won by code-first approach, at least in my opinion, and that’s also what Claude Dashboard will need to go towards, or they don’t care about deep BI integration, as BI tools will probably do it better integrated with agnostic agents anyway, so instead, they just try to capture the top of the BI market.

I wrote more about BI is Dead on AI Reveals Why BI Still Matters, about BI-as-Code and the New Era of GenBI and Has Self-Serve BI Finally Arrived Thanks to AI?

Also, usually a rule of thumb, at a tool like Claude, that pretends to do it all, is OK if you want to move fast and automate things, but not if you’d like to maintain, or improve, or craft something beautiful or truly important to your business, like Claude Design and others.

I still think a dedicated tool is needed, and I use both today: Claude Code (not design), and then ask it to create an SVG, so I can add and edit it myself. With something like internal dashboards and KPIs, that’s the same. If the insights are truly important, none will just one-shot them and then forget, they need care over time. That’s where I see dedicated BI Tools still winning. Even better, if you use open-source Code-first BI tools, you have the best parts of both.

# Further Reads

• See BI Tools categorized in BI-as-Code and others
• Vibe Coding dashboards.

---

Origin: Build live dashboards and animate explainers with Claude | Claude by Anthropic
References: Claude Code, Self-Service BI