Claude Data is Anthropic’s connected-data workspace for turning company questions into dashboards, data stories, narrated short videos, and written analysis. It was previously called Claude Dashboards: the dashboard beta launched in October 2026, then became generally available under the new name on November 10. Existing dashboards continue to work. To create one, connect a data source, ask Claude a specific question from chat or Artifacts > Data, then inspect the SQL and refresh time behind each chart before sharing.
An analytics commenter asked whether people return to generated dashboards after the first view, noting that “someone has to keep the data fresh and notice when something breaks.” That is the right question for any dashboard. Claude can draft and update a view; your team still needs clear metric definitions, a refresh expectation, and an owner for decisions. The Claude Data Help Center documents the current workflow and its limits.

Key takeaways
Claude Data is the new name for Claude Dashboards, which moved from beta to general availability on November 10, 2026. Existing dashboards keep working.
Connect a supported warehouse or an app already connected to Claude, then ask for the business question you want answered. You can start in chat or in Artifacts > Data.
Each chart exposes its SQL query and last refresh time. Check that the query matches your metric before using the number.
Certified queries and live sharing are beta features. Live sharing can show viewers results run through the sharing editor’s data access, so inspect the dashboard before enabling it.
Claude Data is suited to quick exploration and sharing results. Anthropic recommends another tool when a question requires deeper analysis.
What Claude Data does
Claude Data connects a question about company information to a data source, generates a SQL query for each chart, and displays the results. You can ask for a dashboard, a data story, visual analysis, or a short narrated video. The output is saved in the Artifacts tab. A chart shows the query Claude used and when the data was last refreshed, so a reviewer can inspect the logic behind a number.
The product is available on Pro, Max, Ultra, Team, and Enterprise, on web, desktop, and mobile. Claude Data counts toward plan usage. The current help guide lists data warehouses such as Amazon Redshift, BigQuery, ClickHouse, Databricks, and Snowflake, plus other apps already connected to Claude; Salesforce opportunities is one example. Availability still depends on your organization’s connector setup and permissions.
| Choose Claude Data for… | Use a different reporting path when… |
|---|---|
| A quick question about a connected company data source. | The result must become a durable, scheduled report with an established production owner. |
| A first dashboard or data story to inspect with a small group. | The audience needs carefully managed row-level access or a deployment process your BI system already provides. |
| Exploring a trend and checking the SQL behind each chart. | The work needs a deeper analytic model, repeatable testing, or a broader transformation pipeline. |
| Sharing a specific set of current query results. | The dashboard’s source access, metric definitions, and viewers have not been reviewed. |
Anthropic describes Claude Data as a place for quick, exploratory questions and recommends continuing deeper analysis in another tool when needed. That gives it a clear role: use it to investigate a question and produce a reviewable artifact; do not assume every exploratory view is ready to become an organization’s reporting system.
Claude Data vs. Sheets, Whiteboards, and Meetings
These Claude surfaces solve different jobs. The Anthropic launch note introduced them together but does not make them interchangeable.
| Claude feature | Main work | A useful starting point |
|---|---|---|
| Claude Data | Explore connected company data and make dashboards or data stories | “How did weekly signups by plan change over the last six months?” |
| Claude Sheets | Create and edit spreadsheet tabs, formulas, formatting, and drop-downs | “Build a hiring budget with these assumptions.” |
| Claude Whiteboards | Arrange ideas visually with teammates | “Map the open decisions and owners for this launch.” |
| Claude Meetings | Work during a Google Meet and continue from meeting notes or artifacts | “Capture the decision and draft an action list during this call.” |
Use Claude Data when the question depends on an authorized source of company data. Use Sheets for a grid of editable values and formulas. Use Whiteboards for shared visual thinking. Meetings is a separate live-call workflow; none of the other three features joins a call.
How to build and check a dashboard
1. Connect the source you want to ask about
Before creating a dashboard, connect a warehouse or an app through Claude’s connector setup. The official guide names Redshift, BigQuery, ClickHouse, Databricks, and Snowflake as warehouse examples. It also says you can use other apps already connected to Claude, such as Salesforce. Choose the source that holds the data you intend to use; Claude Data is not a way to query an arbitrary database without a connector.
For a team, confirm that the connector is enabled for your organization and that you have the needed permission. If the source contains multiple tables or metric definitions, identify the relevant table or certified definition before you prompt. This reduces ambiguity and gives a reviewer a concrete place to start.
2. Ask the business question, not just for a chart
Start from a chat by describing the question. Or open Artifacts, select Data, and enter the request there. Anthropic’s examples include revenue by region this quarter and weekly signups over six months split by plan.
A more reviewable prompt can name the source, time window, metric, grouping, and intended audience:
Using the connected analytics warehouse, compare weekly new signups for the last six complete months by plan. Use the approved signup definition if one is available. Show the trend as a dashboard, include the date range and source table, and flag any week with missing data. Do not infer why a change happened.This asks for a specific analysis while leaving the result open to inspection. If you know the table or app, name it. Ask Claude to show how it defines the metric rather than assuming that “signups,” “revenue,” or “active customer” means the same thing across every team.
3. Refine the view and inspect its query
After the first result, you can ask Claude for changes in the chat, such as adding a region filter or changing from daily to weekly. You can also click a chart, title, or caption to edit it directly. Each chart exposes its SQL query; open it and check the joins, filters, date range, and aggregation against your intended question.
If the query does not count what you meant, tell Claude which definition or condition is wrong. Do not correct only the chart title while leaving a mismatched query underneath. A chart can look plausible and still answer a different question. When the number matters, ask a data teammate to review the query and the source table.
A Power BI commenter summarized the issue in the earlier Claude Dashboards beta discussion: “Security, governance, consistent KPI definitions, and long-term maintenance still matter.” That comment is not an audit of the current product, but the principle applies to generated analytics: check the definition and query, not just the visual.
4. Check freshness and version history
Claude Data shows when chart data was last refreshed. The Help Center says a dashboard refreshes its data when someone opens it and the data is due for refresh. It does not describe a fixed scheduled refresh. “Refresh data” re-queries based on a specified interval and user access patterns; Anthropic also says caching and per-dashboard refresh cadences are supported.
If other people will rely on the dashboard, agree on how fresh the numbers need to be and who will notice an unexpected result. Version history includes direct edits and lets an editor preview or restore an earlier version. A restore changes the dashboard for everyone who opens it, and the action appears in history. View-only users cannot edit or restore a version.
5. Share only after reviewing what viewers will see
Everything you create starts private to you. Use Share when you are ready to invite others; Claude Data artifacts can also be shared outside your organization. Normal sharing and live sharing are different. Without live sharing, each viewer loads results through their own connected sources. With live sharing—currently beta—people in your organization can see the approved dashboard results through the sharing editor’s access, even if they do not have access to the underlying connector. They cannot use that access to run other queries. The shared scope is limited to the dashboard’s exact queries, which cannot be changed without editor approval.
Before turning it on, inspect the dashboard as if you were a viewer: check each chart, its query, the selected date range, and any sensitive grouping. Ask whether everyone who can open the artifact should see those results. Live sharing is convenient when viewers need the same approved snapshot; it also changes whose connector access supplies the data.
Which Claude Data output fits the question?
| Output | Good for | Review before sharing |
|---|---|---|
| Dashboard | Several charts for an exploratory question | Query, source, filters, last refresh time, and audience |
| Data story | A written narrative built from metrics | Whether each statement is supported by the displayed numbers |
| Visual analysis | A written summary of results | Definitions, caveats, and whether the summary separates data from interpretation |
| Narrated short video | A short walkthrough of numbers | The exact charts and claims shown to viewers |
For example, a sales lead might ask for revenue by region and compare the current quarter with the prior quarter. The dashboard can make the first comparison visible; it cannot decide whether a change came from seasonality, a pricing change, or missing data unless the evidence and analysis support that explanation. Add the relevant source and definitions, inspect the SQL, then use a deeper analysis workflow if the decision requires a causal account.
Certified queries and admin settings
Certified queries are in beta. A data team can approve a query definition for Claude to reuse when creating dashboards. Charts that use a certified query show a badge. The Help Center documents a flow to inspect or edit the query, add instructions, test it, and publish it. Editing a certified query removes its certification until it is checked and published again.
Treat certification as shared metric governance, not as a visual decoration. Agree on the metric meaning and query owner first. The documentation allows anyone with write access to certify a query, so organizations should decide who is responsible for that action. A badge communicates that a query was certified; it does not make every chart or conclusion correct by itself.
On Team, Claude Data and both beta features are on by default. As of November 16, 2026, Enterprise Claude Data is scheduled to turn on by default November 20 unless an owner has already saved a setting; certified queries and live sharing are off by default. Owners manage Claude Data under Organization settings > Artifacts. Enterprise owners can use custom roles to limit access to groups. If you do not see Data in Artifacts, ask an owner to check the organization setting and connector access.
Where Tabbit fits around Claude Data
Tabbit Browser is not a Claude Data connector, warehouse, dashboard, or governance layer, and no direct integration was verified. Its adjacent role is to collect reference pages—such as metric definitions or product documentation—before an analytics discussion. The researcher browser guide and deep research workflows cover that separate task. Keep the source links with any definition you bring into a dashboard review.
If those references are spread across many tabs, see browser productivity, profiles, workspaces, and tab groups, and Tabbit Browser practices. For comparing source pages, read research with browser agents and Claude Cowork browser workflows. The AI browser comparison is useful when choosing a browser for that research work.
A practical decision
Claude Data is a good next step when you have a connected source, a clearly worded question, and someone ready to review the query and sharing scope. It is less suitable as an unattended reporting system when metric ownership, refresh expectations, or audience permissions have not been decided. Start with one exploratory question, confirm that the SQL answers it, then decide whether the result should be shared as a dashboard, certified query, or handed off to your established analytics tool.
FAQ
What is Claude Data?
Claude Data is Anthropic’s connected-data workspace for creating dashboards, data stories, narrated short videos, and written analysis from questions about company data. It was previously called Claude Dashboards.
How do I create a Claude Data dashboard?
Connect a warehouse or app to Claude, then ask a question in a chat or open Artifacts > Data and describe the analysis. Name the source or table when you know it, and ask the business question rather than only requesting a chart type.
Does Claude Data show the SQL behind a chart?
Yes. Each chart shows the SQL query Claude wrote and the last data refresh time. You can inspect or edit the query and tell Claude what to change if it does not match the metric you intended.
How often does Claude Data refresh dashboards?
The Help Center says data refreshes when someone opens a dashboard and it is due for refresh. It does not describe a fixed scheduled refresh; the refresh action re-queries based on the configured interval and access patterns.
What are certified queries in Claude Data?
Certified queries are approved SQL definitions that Claude can reuse when it builds dashboards. The feature is in beta, and certification is a governance process; a certified badge does not replace checking what a metric means.
Can people see a shared dashboard without access to its connector?
With live sharing enabled, viewers in your organization can see the dashboard’s approved query results through the sharing editor’s access, even if they do not have access to the underlying connector. Review the dashboard contents and settings before enabling it.