Tableau MCP

Tableau's MCP server is a vendor-maintained bridge between an AI assistant and a Tableau Server or Tableau Cloud site. It gives the agent two things that matter for answering analytics questions correctly: a way to discover what exists, listing published data sources, workbooks, views and their metadata, and a way to query, running headless requests against a published data source so it gets aggregated numbers back rather than a rendered chart image. That distinction is the point. Instead of inventing a field name, the agent reads the actual columns, their data types and the descriptions curated by the analytics team, then builds a query against the semantic layer that already encodes the organization's own definitions of revenue, churn or active user. The result inherits the governance living in the published data source, including row level security tied to the credentials in use, so an agent cannot surface numbers a person is not entitled to see. It fits analytics engineering and BI teams that have invested in curated Tableau content and want conversational access on top of it without exporting extracts into a separate system. The project is open source and runs locally against a site's credentials, which lets a team restrict it to specific data sources before opening it up to a wider group of users.