Claude + Tableau: What the Integration Can (and Can't) Do in 2026
Yes — Tableau ships an official MCP server, and it’s one of the best-maintained in analytics. The tableau/tableau-mcp repo lives in Tableau’s own GitHub org (Apache-2.0, 43 releases, v2.24.0 shipped July 8, 2026), and there’s a hosted remote endpoint at mcp.tableau.com for Tableau Cloud customers — OAuth 2.1 against your Tableau identity, respecting your per-user permissions. Tableau’s own developer site pairs it with Claude. It isn’t a one-click listing in Anthropic’s connectors directory — you add it as a custom connector — but the server is unambiguously official. Once wired up, Claude can query datasources, search workbooks, and pull Pulse metrics in plain English. The catch that applies to every Claude MCP setup: it runs when you ask or on a schedule you set, never because something happened in Tableau.
Here’s exactly what the integration does, how to set it up, where the limits bite, and what to use if you want Tableau work that runs on its own.
What the Tableau MCP server does
The official docs list a genuinely useful tool set:
- Query datasources — the query-datasource tool calls the VizQL Data Service for field-level queries with filters and aggregations. “What was West-region revenue by month this year?” becomes a governed query, not a screenshot guess.
- Read metadata — field names, types, and structure via VizQL Data Service and the Metadata API, so Claude knows what it’s querying.
- Find and read content — workbook and view tools including search-content across your site.
- Work with Pulse — metric definitions, metrics, subscriptions, and insight bundles (Tableau Cloud’s AI metrics layer).
This lines up with where Tableau is heading: at TC26 (May 2026) it positioned MCP as the “universal translator” for headless analytics inside Slack, Teams, and Claude.
How to set it up
Tableau Cloud (the easy path):
- In Claude (web or Desktop), go to Customize → Connectors → Add custom connector and enter
https://mcp.tableau.com. Remote custom connectors require a paid Claude plan. - Sign in via OAuth 2.1 with your Tableau Cloud account. Claude gets exactly your permissions — nothing more.
- Ask away: “find the churn dashboard and summarize what changed this week.”
Tableau Server (self-hosted): the hosted service covers Cloud only, so run the MCP server yourself via npx or Docker with a Personal Access Token in env vars (PAT_NAME, PAT_VALUE), attached to Claude Desktop or Claude Code as a local server.
The limits that actually matter
- Cloud gets the hosted endpoint; Server doesn’t. On-prem Tableau Server means self-hosting the MCP server and managing PATs — note Tableau Cloud PATs expire after ~15 days of non-use, which quietly kills forgotten setups.
- No triggers, no monitoring. The MCP server runs when you ask or on a timer, and nothing in Tableau can start it. Tableau actually has real webhooks (DatasourceRefreshFailed, WorkbookRefreshFailed and more), but Claude can’t receive them. An extract can fail at 2am and Claude won’t know until you ask or a scheduled task looks.
- Approval-gated delivery. A scheduled task can summarize a view every Monday, but each email to leadership waits on your approval, and nothing reacts when the data changes.
- Scheduled, not event-driven. The closest thing to “running on its own” is Claude Cowork’s scheduled tasks, which run hourly, daily or weekly on paid plans, not on inbox events.
So Claude is great for “interrogate the data with me right now” — and not built for “flag failed refreshes as they happen and retry them.”
If you want Tableau work that runs on its own: Carly
The moment you want something to happen around Tableau without you in the loop (the refresh-failure alert, the Monday brief sent to leadership, the board deck filed before the meeting), you’ve crossed past what the MCP server is for.
That’s where Carly fits. Carly is an AI executive assistant built to act on triggers. Tableau connects to Carly via your own API credentials — create a Personal Access Token and paste it on carlyassistant.com/integrations; Carly drives the full Tableau REST API like any native connection:
- Monday 7am KPI brief — Carly pulls the exec dashboard as a high-res PNG plus the underlying CSV (Query View Image / Query View Data), writes the week-over-week story, and emails both to leadership before standup.
- Refresh-failure triage — Tableau’s DatasourceRefreshFailed webhook hits Carly, which alerts the data owner with context and can fire
runNowon the extract refresh task to retry once. - Board-deck automation — the night before the board meeting, Carly downloads the workbook as PowerPoint, drafts speaker notes from the underlying data, files the deck to Drive, and sends you the summary.
- Per-region digests — one dashboard, N emails: Carly loops
vf_Region=view filters over the image and data endpoints so each VP gets their own numbers with a written narrative. - No-code setup — describe the workflow in plain English; Carly interviews you and builds it.
- Actually sends — drafts and sends email across Gmail and Outlook, updates records, manages tasks.
- Connects to anything — 200+ native integrations, plus any other tool via your own API key.
AI agents start at $35/month, and steps in a workflow that don’t use AI run free.
Claude vs Carly
| Claude (Tableau MCP) | Carly | |
|---|---|---|
| Query datasources in plain English | Yes (VizQL Data Service) | Yes, via REST API |
| Search workbooks & views | Yes | Yes |
| Reacts to refresh failures (webhooks) | No | Yes |
| Scheduled exports (PNG/CSV/PDF/PPTX) | Partly (a scheduled task can pull view images and data) | Yes |
| Trigger extract refresh (runNow) | No | Yes |
| Works while laptop is closed | Scheduled tasks only | Yes (cloud) |
| Emails the numbers to stakeholders | Gmail, with your approval per send | Yes (Gmail + Outlook) |
| Pricing | Pro $20 / Max $100–$200 | AI agents from $35/mo |
Claude with the Tableau MCP server is a strong analyst you interrogate in a chat. Carly is a teammate that ships the numbers and reacts when pipelines break.
Frequently Asked Questions
Does Claude work with Tableau?
Yes. Tableau’s official MCP server connects Claude to your Tableau site — hosted at mcp.tableau.com for Tableau Cloud (OAuth 2.1, add as a custom connector) or self-hosted with a PAT for Tableau Server. Claude can query datasources, read metadata, search content, and work with Pulse metrics. It’s a custom connector, not a one-click directory listing, and it runs when you ask or on a schedule, never on a Tableau event.
Can Claude email me a dashboard every Monday?
Partly. The MCP server has no schedule of its own, but a Claude scheduled task can pull the dashboard every Monday on a paid plan; the email itself waits on your approval. Tableau’s own Subscriptions API can email static views on a schedule, but there’s no narrative and no cross-tool follow-through. For a written brief with the PNG and CSV attached that sends itself, use a trigger-based agent like Carly.
Does the hosted Tableau MCP work with Tableau Server?
No — mcp.tableau.com covers Tableau Cloud only. Server customers self-host the open-source MCP server (npx/Docker) with PAT auth. Watch PAT expiry: Cloud PATs lapse after about 15 days of non-use.
Can Claude fix a failed extract refresh?
Not on its own. Claude can’t receive Tableau’s webhooks, so it won’t know a refresh failed until you ask or a scheduled task checks. The REST API exposes runNow on extract refresh tasks; Carly can catch the DatasourceRefreshFailed webhook, alert the owner, and retry the refresh automatically.
More: Claude + Snowflake · Claude + Power BI · Claude + Databricks · Claude + Salesforce · Claude connectors · Can Claude send emails
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
