Claude Dashboards & Claude Motion
Two new templates landed in the artifact panel on October 8, 2026, and they are not variations on the ones already there. Claude Dashboards reaches outside the conversation into your data warehouse. Claude Motion produces something the other templates cannot: a moving image — and it produces it as code, not as video.
The same announcement quietly ended the beta on Docs, Slides and Design, which are now on every plan including Free. If you read an AILmanac page (or anyone else's guide) written before October 8 saying those three are "paid, beta," that sentence is now wrong.
- Tell Dashboards and Motion apart from the templates you already have, and from a plain artifact
- Know exactly which plan gets which, and what an Enterprise owner has to switch on
- Understand the one mechanism behind each: a SQL query per chart, and an animation compiled from editable code
- Avoid the sharing trap — who sees what when you share a dashboard, and why a connector-backed artifact can never have a public link
- Recognize the three jobs these are the wrong tool for
- Act on the December 14, 2026 Claude Design deadline before it passes
The five templates, at a glance
| Template | What it produces | Plans | Status |
|---|---|---|---|
| Docs | A living, multi-tab document | All, incl. Free | GA |
| Slides | A presentable deck | All, incl. Free | GA |
| Design | Visuals, mockups, prototypes | All, incl. Free | GA |
| Dashboards | Charts driven by live queries against your data | Pro, Max, Team, Enterprise | Beta |
| Motion | A short animation, exported as MP4 | Team, Enterprise | Beta |
On Enterprise, both new templates are off until an owner turns them on in organization settings, under Artifacts — and an owner can narrow access to specific groups with custom roles. On Team, Motion is on by default.
Claude Dashboards
The mechanism, in one sentence
You ask a question in plain language; Claude writes a SQL query for each chart and runs it against your data source. That is the whole product, and understanding it explains every strength and every limit that follows.
Supported data warehouses: Amazon Redshift, BigQuery, ClickHouse, Databricks and Snowflake. It can also read apps you have already connected to Claude, Salesforce being the headline one.
Why the per-chart query matters
Because each chart is backed by a named query rather than a model's recollection, a dashboard is auditable in a way a chat answer is not. Every chart shows its own query and when its data was last refreshed. That turns the usual trust problem inside out: instead of asking "did Claude hallucinate this number," you click the chart and read the SQL that produced it.
This is the single habit worth forming on day one.
- "Build a dashboard of this quarter's pipeline by region" gets you a chart. "I need to decide where to move two reps next quarter — show me what would tell me that" gets you a chart plus the comparison that makes it actionable. Anthropic's own guidance is to say what you want to know, not just what to plot.
- Say where the data lives. Pointing at a specific warehouse table or connected app is the difference between one query and three guesses.
- This is the step people skip. Read the SQL: check the date range, the filter, the join, and whether it counts rows or sums a column. A correct-looking number from the wrong filter is the actual failure mode here, not invention.
- "Add a filter for region," "split new versus renewal," "use fiscal quarters." Claude rewrites the query rather than redrawing a static picture.
- Anthropic's framing is explicit: dashboards are made for quick, exploratory questions. When a question needs real analysis, send the dashboard onward to a tool built for it.
A dashboard brief that produces something you can act on
Build me a dashboard from our Snowflake warehouse (schema: SALES_PROD). The decision I'm making: which two regions get the next headcount. Show me: 1. Closed-won revenue by region, last 4 fiscal quarters, as a trend 2. Win rate by region over the same period 3. Average deal cycle in days by region 4. Open pipeline by region, split new business vs renewal Rules: - Use fiscal quarters, not calendar. - Exclude deals marked is_test = true. - For every chart, label the exact table you queried. - If a field I named doesn't exist, say so and ask — do not substitute a similar-sounding column.
That last rule earns its place. The nastiest dashboard error is not a broken chart — it is a plausible chart built on amount when you meant amount_arr.
Sending a dashboard somewhere else
A dashboard is a starting point, not a home. Anthropic lists onward destinations including Amplitude, Grafana, Hex, Mixpanel, Omni, Perplexity, PostHog and Sigma, with Databricks Genie, Looker, monday.com and Tableau named as coming.
Read that list as a statement of intent: Dashboards is positioned as the fastest way to ask a question, and explicitly not as a replacement for the BI tool where your governed metrics live.
The sharing model — read this before you share anything
Dashboards start private to you. Sharing one makes it an artifact that calls a connector, and connector-backed artifacts follow rules that surprise nearly everybody:
- Queries run as the viewer, not as you. Per Anthropic: "When someone opens your artifact, it connects to their connectors and data sources." Two colleagues can open the same dashboard and legitimately see different numbers.
- A viewer without access sees an error, not your data: "If they don't have access to an underlying data source, that part of the artifact shows an error instead of your data." That is the feature working correctly — it is also why your dashboard looks broken to the person you sent it to.
- There is no public link. "There are no external or public links and no per-person recipient selection. Anyone in your organization who has the link can open the artifact." Live data and no-login external access are mutually exclusive on the same page.
- Connecting apps to an artifact is a web-and-desktop capability on Pro, Max, Team and Enterprise. Don't plan a mobile-first review loop around it.
The practical consequence: a dashboard is an internal instrument, not a client deliverable. If an outside recipient needs the numbers, export the figures into a Doc, a deck, or a generated file and send that.
Claude Motion
Not a video model
This is the fact that most coverage gets wrong, so it is worth stating plainly: Motion does not generate video. Claude writes editable code that animates text, numbers, shapes, images and charts, and the artifact plays like a short video. The direct consequence, in Anthropic's words, is that it cannot create realistic footage or people.
That restriction is also the whole advantage. Because the animation is a program:
- You can change one word, one number or one timing value and everything else stays exactly as it was.
- The same brief, re-rendered after a data correction, is still the same animation — not a new roll of the dice.
- You can direct it in prose ("slow down the second scene") or edit it directly in the editor.
If that mechanism sounds familiar, it is the same idea as the broader code-driven explainer-video pipeline — Motion is the first-party, no-setup version of it.
Using it well
- A sequence, a process, a value changing over time, a walkthrough of steps. If the content is a list, make Slides instead — animating a list is slower to produce and slower to read.
- "30 seconds, for an all-hands, playing on a big screen behind me" and "90 seconds, for new-customer onboarding email" produce legitimately different animations. Attach the source report or data instead of retyping it.
- Ask in any conversation, choose Output then Motion in the message box, or start from a Motion template in the Artifacts tab. Pick explicitly when Claude keeps producing a deck.
- "Slow down the second scene," "make the revenue figure land after the chart finishes drawing." Single changes land reliably; a list of five rarely does.
- Anthropic names Adobe, Descript, HeyGen, Higgsfield, invideo, Luma AI and Runway as tools you can open the output in, with Canva and Captions listed as coming. Voiceover, captions and music are not documented as Motion features — plan to add them downstream.
Brief an animation that doesn't need a second attempt
Turn the attached Q3 report into a 30-second animated explainer. Audience: the whole company at an all-hands, on a projector behind a speaker. It plays once, with no narration from the file itself. Structure, four scenes: 1. The one number that matters (revenue, with last quarter for contrast) 2. Where the growth came from — animate the regional chart drawing in 3. The one thing that got worse, stated plainly 4. What we're doing about it next quarter, three short lines Constraints: - Readable from the back of a room: large type, high contrast, no more than seven words on screen at once. - No decorative motion. Every movement should mean something changed. - Use only figures from the attached report. If a number I asked for isn't in it, leave the slot empty and tell me.
When these are the wrong tool
An honest reference says what a feature is bad at. Three cases:
| You want… | Don't use | Use instead |
|---|---|---|
| A governed metric your whole company reports on | Dashboards — it answers questions, it does not define metrics | Your BI layer; send the dashboard there |
| A live data page a client opens without logging in | Dashboards — connector-backed artifacts have no public link | Export the figures into a Doc or deck, or a generated file |
| Footage, a presenter, a talking head | Motion — there is no video model behind it | A video tool; Motion's MP4 can be a component inside it |
And one cost note that applies to both: there is no separate price. Dashboards and Motion count toward your plan's usage limits like anything else, and Anthropic flags that longer or more complex animations use more of your allowance. The plans and billing page has the general shape.
A deadline worth putting in your calendar
The October 8 changes came with a sunset. The standalone Claude Design site closes on December 14, 2026, after which claude.ai/design simply takes you to Claude.
If your team ever used it, this is a real to-do with a date on it:
- Open Artifacts in Claude and click "Migrate team design systems," from the banner or the Claude Design menu. Each migrated design system keeps its owner and the people it was shared with. Anyone who can use Claude Design can start the migration — it is not owner-only.
- Empty design systems and ones built from a built-in starter theme are not migrated. Neither are your chats with Claude, nor comments on projects and design systems — those stay in the standalone version and are deleted when it closes.
- This is the irreversible part. Public links to projects in the standalone version also stop working after it closes.
- A published, non-private design system becomes visible organization-wide after migrating. Review which ones that applies to before you click.
- Artifacts must be enabled, and on Enterprise design systems are off by default until an administrator turns them on. Do that before the migration, not after.
Terms worth knowing
Quiz
Check yourself
0/4- October 8, 2026: Dashboards and Motion arrived in beta; Docs, Slides and Design left beta and reached every plan including Free. Any guide still calling those three 'paid, beta' is out of date.
- Dashboards writes a SQL query per chart. Open the query before you trust the number — the real failure mode is a correct-looking figure from the wrong filter or column.
- Shared dashboards run as the viewer. Different people see different data, and a viewer without access sees an error — by design.
- No connector-backed artifact can have a public link. A dashboard is an internal instrument; for external recipients, export.
- Motion is code, not video: endlessly editable and perfectly repeatable, but it can never produce footage or people. No voiceover or captions documented — add those downstream.
- Anthropic scopes Dashboards to quick, exploratory questions and names onward destinations like Grafana, Hex and Sigma. Treat it as the fastest way to ask, not as your BI layer.
- Calendar item: the standalone Claude Design site closes December 14, 2026. Migrate design systems from the Artifacts page, and save chats and comments first — those are deleted.
Sources & further reading
- Anthropic — Build live dashboards and animate explainers with Claude (announcement, October 8, 2026; connector and onward-destination lists)
- Anthropic Help Center — What are artifacts and how do I use them? (the five templates and their exact plan availability)
- Anthropic Help Center — Get started with Claude Dashboards (creation, per-chart queries, sharing, scope and usage limits)
- Anthropic Help Center — Get started with Claude Motion (creation, editing, MP4 export, plan and org settings)
- Anthropic Help Center — Use live artifacts (whose connections run on view, and why there is no public link)
- Anthropic Help Center — Migrate from standalone Claude Design to Claude (the December 14, 2026 closure and what is not migrated)
Next
- Artifacts: Live, Runnable Outputs — the panel all five templates render into
- One Claude: The Merge, Claude Docs & Claude Slides — the three GA templates in depth, and their limits
- Connectors — wiring the data sources Dashboards queries
- Data Analysis playbook — the analysis work a dashboard hands off to