Everyday AI Guides · AI for Work & Productivity
Claude Dashboards lets eligible Claude users turn questions about connected business data into visual dashboards. You can ask for revenue trends, signups or pipeline breakdowns, inspect the SQL query behind each chart, and check when the data was last refreshed. This guide shows how to set one up, write a useful request and verify the results before anyone relies on them.
Reviewed October 10, 2026 · Independent educational guide · No hands-on product testing claimed
What Are Claude Dashboards?
Claude Dashboards is a dashboard template within Claude Artifacts. Instead of manually building every chart, you describe the business question and Claude generates queries and a visual display from the data source you have connected. Each chart provides a way to examine the query used to produce its numbers.
Anthropic introduced the feature in beta on October 8, 2026. Its documentation describes it as a way to answer quick, exploratory data questions rather than a replacement for every reporting or business-intelligence workflow. Read the official announcement.
What “live” means here: The dashboard works with a connected source and displays a last-refreshed indication. Do not assume continuous, second-by-second updates. Always check the displayed refresh time and source before describing a result as current.
Good starting questions
How did paid revenue change month to month? Which regions produced the most orders? What share of open opportunities belongs to each stage?
Questions needing more control
Audited financial statements, regulated decisions, complex forecasts or official company KPIs may need additional review and a governed reporting system.
How to Use Claude Dashboards: Step by Step
Use this sequence if your Claude plan supports the Dashboards template and you have permission to work with the underlying data.
Confirm your plan and capabilities
Sign in to Claude with a Pro, Max, Team or Enterprise account. Check that cloud code execution and file creation are enabled under the relevant Capabilities settings. Enterprise owners may also need to enable Dashboards under Organization settings → Artifacts.
Connect an approved data source
Use an authorized data warehouse or supported app connection. Confirm the account can read the relevant tables and that your organization allows the data to be used in Claude. Do not paste credentials or private customer records into a prompt.
Open the Dashboards template
Go to Artifacts and select the Dashboards template. Alternatively, ask for a dashboard in a chat, or select Output → Dashboards when that control is available in your interface.
State the business question and metric rules
Name the intended source, time zone, date range, filters and metric definitions. For example, specify whether refunds and pending transactions count toward revenue.
Refine, inspect and verify
Request an additional filter or a weekly breakdown. Then open each chart’s underlying query, compare its totals against the source and record when the data was refreshed. Only share the dashboard after that review.
For instructions that are easier to reuse across work tasks, try the free AI Prompt Generator and add clear inputs, exclusions and expected output requirements.
Claude Dashboards Data Sources: What Can You Connect?
Anthropic names Amazon Redshift, BigQuery, ClickHouse, Databricks and Snowflake as example supported data platforms. It also describes using connected business apps, including Salesforce, when the relevant connection is available and authorized.
| Data location | Example task | What to confirm |
|---|---|---|
| Data warehouse | Orders by month and region | Table name, field types, query permissions and refresh cadence |
| Connected CRM | Open opportunities by stage | Connector access, owner permissions and meaning of each stage |
| Another supported app | Operational trend or category breakdown | Which records are accessible and whether fields are complete |
Important distinction: Uploading a CSV file to an ordinary chat is not the same as connecting a supported source to a Claude Dashboard. To follow the connected-data workflow below, put the fictional sample records into a table in a supported, authorized data source first, or adapt the exercise to an existing approved dataset.
Claude Dashboards Example: A Sales Report You Can Check
To make the tutorial practical without publishing anyone’s real business information, use this fictional eight-order dataset. It demonstrates why the definition of a metric matters more than the appearance of its chart.
Assume a table named demo_orders with the following records. The dates use YYYY-MM-DD, the amounts are in U.S. dollars, and the exercise deliberately includes a refunded order and a pending order.
| Order | Date | Region | Channel | Status | Amount |
|---|---|---|---|---|---|
| 101 | 2026-09-02 | East | Organic | paid | $120 |
| 102 | 2026-09-05 | West | paid | $80 | |
| 103 | 2026-09-08 | East | refunded | $50 | |
| 104 | 2026-09-14 | West | Organic | paid | $150 |
| 105 | 2026-10-02 | East | Organic | paid | $200 |
| 106 | 2026-10-04 | West | paid | $90 | |
| 107 | 2026-10-07 | East | pending | $75 | |
| 108 | 2026-10-08 | West | Organic | paid | $160 |
Define “paid revenue” before making the chart
For this exercise, paid revenue = the sum of amount for rows where status is exactly paid. Exclude pending and refunded rows. Do not use this example as a universal accounting definition; real businesses may handle refunds, taxes, currency conversion and recognition differently.
Illustration drawn for this guide; not a screenshot or output from a tested Claude account.
Additional cross-checks: East has $320 in paid revenue, West has $480, Organic has $630, and Email has $170. Each pair should sum to $800. October is about 28.6% higher than September in this small sample; it is not evidence of real-world growth.
Why this catches errors: If a chart reports $925, it has included every order amount rather than only paid rows. If it shows eight paid orders, it ignored the status field. These errors can survive even when the dashboard looks polished.
Claude Dashboards Prompts: Build, Fix and Verify
These are reusable starting prompts. Adjust table names, available fields and permissions to match your own data source. Do not treat an AI-generated query as validated SQL until you inspect it.
Build a Claude Dashboard using my authorized demo_orders table. Business question: How does paid revenue compare between September and October 2026, and how is it split by region and channel? Metric rules: - Paid revenue means SUM(amount_usd) for status = 'paid' only. - Exclude pending and refunded rows. - Use the order_date field, not the system's current date. - Show USD amounts and use the same filter rules for every chart. Create monthly paid-revenue bars, paid-order count, revenue by region, and revenue by channel. For every chart, make it easy to inspect the source, SQL query, filters and last refresh time. Do not assume a missing field exists. Ask me to confirm any ambiguous schema or metric definition before proceeding.
The paid-revenue chart does not match my source data. Review the exact query used by this chart and explain: 1. Which table and columns it reads. 2. Which statuses it includes or excludes. 3. Whether dates are grouped correctly and in the intended time zone. 4. Whether joins could duplicate rows. 5. Whether refunds, nulls or currency fields change the total. Propose a corrected query for my actual database dialect. Show the revised metric definition and explain what changed. Do not invent missing tables, columns or validation results.
Prepare an audit checklist for this dashboard before I share it. For every metric list: the source, definition, SQL query, date range, filters, aggregation method, units, and last refresh time. Mark each item as: - CHECKED AGAINST SOURCE - REQUIRES HUMAN VERIFICATION - UNAVAILABLE Flag contradictory totals, ambiguous definitions, incomplete data, and privacy or sharing risks. Do not label any item verified unless there is actual evidence from the accessible source. Finish with the five highest-priority manual checks.
How to Verify Claude Dashboard Charts and SQL Queries
A query is an explanation of how a number was calculated, not a guarantee that it is the right number. Use this checklist on each important chart:
- Source: Is this the intended connection, database and table?
- Metric: Does “revenue,” “customer” or “conversion” have a written definition?
- Filters: Are refunded, canceled, test and pending records handled correctly?
- Dates: Do date ranges, grouping and time zones match the question?
- Joins: Could a join create duplicate orders or customers?
- Units: Are amounts, currencies, percentages and axis labels clear?
- Reconciliation: Do chart totals match an independent source calculation?
- Freshness: Does the displayed last-refreshed time meet the need?
- Access: Can intended viewers access the required source without exposing sensitive data?
A quick independent check
For the fictional dataset, manually add $120 + $80 + $150 + $200 + $90 + $160. The total is $800. Then verify the two monthly subtotals add up to the same amount. If the dashboard disagrees, inspect its query and source instead of requesting a prettier chart.
If you later place findings into a slide deck, review the numbers and visual labels again using the AI Presentation Maker Checklist.
Claude Dashboards Pricing, Plans and Missing Access
As of October 10, 2026, Anthropic lists the Dashboards template in beta for Pro, Max, Team and Enterprise. It does not list the Dashboards template for Free accounts. Access may also depend on rollout, plan usage limits and organization settings. Check Anthropic’s current plan and artifact documentation before paying for access.
Claude Dashboards not showing? Check these items
- Plan: Confirm you are signed into the expected eligible account.
- Capabilities: Enable code execution and file creation if your account allows it.
- Entry point: Look for Dashboards in Artifacts, or use the Output menu where available.
- Organization restrictions: Ask your Team or Enterprise administrator whether Artifacts and Dashboards are enabled for your role. Enterprise Dashboards are off by default.
- Data access: Confirm the required connector and read permissions exist.
- Still unavailable: Consult the official help center rather than assuming a hidden setting or an unofficial workaround will unlock access.
Cost caution: Dashboards uses your Claude plan’s usage allowance, and connected services may have their own costs or permissions. This is not a promise of unlimited or free dashboard generation.
Sharing Claude Dashboards Safely
Dashboards begin private. You can open the artifact’s Share dialog, but sharing is not equivalent to giving somebody unrestricted access to your connected database. According to Anthropic, viewers use their own connected-app permissions; data-dependent sections can fail if the recipient cannot access the source.
Sharing options vary by plan and organization settings. Anthropic also states that “Anyone with the link” is unavailable for artifacts that use connected apps or ask Claude questions. Do not promise a public live dashboard link for connected-data reports. Review the official artifact-sharing guidance for your account.
- Share only with people authorized to see the underlying information.
- Check whether stored information or chart labels expose sensitive details.
- Confirm the recipient’s data-source permissions rather than testing access by widening sharing unnecessarily.
- For a public presentation, prepare a separate approved summary with sensitive details removed.
Use the broader AI privacy checklist before connecting workplace data, uploading reports or sharing derived results.
Claude Dashboards vs. a Spreadsheet or BI Tool
| Your need | Reasonable starting point | Watch out for |
|---|---|---|
| Ask a quick question about authorized connected data | Claude Dashboards | Incorrect metric interpretation, refresh time and permissions |
| Explore a tiny local sample manually | Spreadsheet | Manual updates and inconsistent formulas |
| Run governed recurring executive reporting | Established BI workflow | Semantic definitions, access governance, audit trail and production reliability |
| Create a client-facing presentation | Verified charts followed by a presentation workflow | Export fidelity, context, source disclosure and confidentiality |
Claude Dashboards can speed up exploratory reporting, but the right tool depends on the consequences of a wrong answer and how much control your organization needs. If your next step is explaining the results in a meeting, the Claude Slides guide covers presentation drafting and verification.
Claude Dashboards FAQ
Is Claude Dashboards free?
No. The Dashboards template is currently listed for Pro, Max, Team and Enterprise plans, not Free. General Claude artifacts are available on more plans, but that does not make the Dashboards template free.
Can Claude Dashboards connect to Snowflake or BigQuery?
Yes. Anthropic lists both among example data warehouse connections, alongside Redshift, ClickHouse and Databricks. Actual access depends on configuration and permissions.
Can I build a Claude Dashboard from a CSV?
A CSV uploaded to a normal Claude chat should not be confused with the connected-data Dashboards workflow. For a connected dashboard, use a supported authorized source; confirm current import or connection capabilities before relying on a CSV-only workflow.
Does Claude Dashboards show its SQL queries?
Yes. Anthropic says each chart exposes the SQL query used to create it. You should still inspect the table, filters, joins and definitions, and compare important totals against the source.
Are Claude Dashboards updated in real time?
They can work with connected data and show when data was last refreshed. That does not establish a universal, continuous real-time update frequency. Verify the source and actual refresh time.
Why can’t I see Claude Dashboards?
Common checks include your plan, feature rollout, code-execution capabilities, Artifacts settings and organization permissions. On Enterprise, the Dashboards template is off by default until an owner enables it.
Can someone without my database access see a shared dashboard?
Do not assume so. Anthropic says artifacts using connected apps rely on each viewer’s own access, and connected-data artifacts have special sharing restrictions.
Final Check: Is Your Dashboard Ready to Use?
Before citing a chart in a meeting or report, confirm five things: the source is authorized, the metric is defined, the calculation matches the records, the refresh time is appropriate, and the audience has permission to view it.
Start with the eight fictional orders above. If the dashboard shows $800 in paid revenue, six paid orders, and the correct monthly split, you have a useful first accuracy check—not proof that every future dashboard will be correct.
Explore more AI for Work & Productivity guides or use our AI Prompt Generator to improve your next dashboard request.
Official Sources and Editorial Notes
- Claude Help Center: Get Started With Claude Dashboards — setup, data sources, SQL query inspection and access.
- Claude Help Center: Artifacts and Plan Availability — template eligibility and capabilities settings.
- Claude Help Center: Publish and Share Artifacts — permissions and connected-app sharing limits.
- Anthropic: Build Live Dashboards and Animate Explainers — October 8, 2026 announcement.
Editorial disclosure: This independent educational article is not affiliated with or endorsed by Anthropic. The practice records and illustrative chart are original fictional teaching examples. They are not claimed results from testing a live Claude account. Product features, availability and permissions can change; confirm them in official documentation. This guide does not provide accounting, security or legal assurance.
