Feature Request: Enable Targeted Sheet Retrieval for Multi-Sheet Views

[jackstein21]

Is your feature request related to a problem? Please describe.

Currently, the get-view-data tool retrieves CSV data for a specified view, but it appears limited to returning data from only a single sheet—specifically the first one when ordered alphabetically.

This creates a limitation for complex Tableau Views (dashboards) that contain multiple worksheets. Because the MCP cannot currently see or access the other sheets on a dashboard, the "curated" data provided by analysts is only partially accessible to the LLM. This makes it difficult for the Agent to provide a comprehensive analysis of a dashboard that was designed to be viewed as a cohesive set of multiple data points.

Describe the solution you'd like

I would like to suggest a few potential enhancements (either individually or in combination) to help bridge this gap:

Option A: A metadata discovery tool. A new tool such as get-view-metadata or list-view-sheets that returns the names or IDs of all worksheets contained within a specific view. This would allow the Agent to understand the full scope of data available before requesting it.

and/or

Option B: Targeted data retrieval. Updating the get-view-data tool to accept an optional sheet_name or sheet_id parameter. This would allow the user or the Agent to specify exactly which part of the dashboard they need to query, rather than relying on the alphabetical default.

Additional context

The goal of using this MCP is to transform our curated Tableau dashboards into "active assets" that stakeholders can interact with through a conversational interface. By enabling the tool to navigate multiple sheets within a single view, we move away from manual dashboard navigation and allow the Agent to fully leverage the work analysts have already put into organizing these data views.

[jackstein21]

I’ve done some deeper investigation using the Tableau Metadata API to understand why the get-view-data tool is currently limited to returning only the first alphabetical sheet.

When querying a dashboard and its constituent sheets via GraphQL, I discovered that internal worksheets return a valid GraphQL id, but the luid returns as an empty string. The REST API’s Query View Data endpoint specifically requires a view-id (LUID) to function. Since these internal worksheets aren't assigned an LUID unless the workbook is published with "Show Sheets as Tabs" enabled, the MCP is effectively blocked from accessing any specific sheet data. Consequently, the API defaults to the first sheet it can find within the dashboard container.

This puts me in a difficult position. To make this MCP a truly viable active asset for stakeholders, I need a way to access all curated data on a dashboard without forcing analysts to change their publishing practices (as enabling tabs often clutters the UI for end-users).

---

Questions for the Maintainers:

1. Is there a known way within the MCP to handle discovery for hidden sheets? Maybe sheet name, or the underlying sheet ID?
2. If this is a hard upstream limitation of the Tableau REST API for non-tabbed views, is there a recommended "Agent-Ready" publishing standard? Alternatively, could the MCP leverage the Metadata API to bridge this gap?

I’m eager to help find a solution that allows the Agent to fully leverage the work analysts have already put into these multi-sheet views.

[anyoung-tableau]

Hey @jackstein21, yeah, the limitation currently comes from the Tableau REST API which unfortunately only returns data for the first sheet. The team behind VizQL Data Service (that powers the query-datasource tool) is currently working to add support for data sources embedded in workbooks which shouldn't have similar limitations. I don't have any timelines I can share just yet but once it is ready, the query-datasource tool should be able to provide those remaining workbook data to the LLM. Stay tuned!

[anyoung-tableau]

I have also been experimenting with other approaches to return the full underlying data from a dashboard, but it's a bit hacky. The tool instead launches Chrome (hidden) to a local page that embeds the dashboard using the Embedding API, then uses the Embedding API's export functions on each sheet and merges all the data together. It also only works when AUTH is direct-trust.

Please let me know your thoughts. If this sounds like it might fit the bill for you, please let us know and we can revisit this approach. Here is the prototype commit: d1f8bee