Claude Live Artifacts + Tableau Integration

Priority: P2 Category: PREP

Executive Summary

Claude Cowork now supports "Live Artifacts" — dashboards and trackers connected to apps/files that auto-refresh with current data when opened. Tableau has an official MCP server that enables Claude to query published data sources, discover schemas, and interact with workbooks. We need to evaluate how to connect our existing Tableau infrastructure to Live Artifacts for real-time reporting within Claude.

Research Phase

Current State

• RBD uses Tableau extensively for reporting and visualization
• Claude Code already has MCP server support configured locally
• Claude Cowork (web) now supports Live Artifacts with MCP connectors
• Tableau has an official MCP server: https://github.com/tableau/tableau-mcp

What Live Artifacts Can Do

• Build dashboards/trackers that refresh with current data on open
• Connect to external services via MCP (Model Context Protocol)
• Available on Claude.ai, Cowork, and Claude Desktop
• Custom connectors via remote MCP available on Free, Pro, Max, Team, Enterprise plans

Tableau MCP Server Capabilities

• Schema discovery — browse published data sources, workbooks
• Live querying — query data from published data sources in real-time
• Administrative ops — explore workbooks, views, permissions
• Requirements: Personal Access Token (PAT) from Tableau Server/Cloud

Constraints

• Need to determine: Are we on Tableau Server or Tableau Cloud?
• PAT generation requires admin access
• Live Artifacts are in Claude Cowork (web), not Claude Code CLI
• Data security considerations — what data should be exposed to Claude?

Options Considered

Option 1: Tableau MCP via Claude Cowork (Web)

Approach: Set up Tableau's official MCP server as a custom connector in Claude.ai/Cowork settings, enabling Live Artifacts to pull from our Tableau data sources directly. Pros: Official integration, auto-refreshing dashboards, shareable with team, no local setup needed per user Cons: Requires Tableau Cloud or publicly-accessible Tableau Server; data passes through Claude's infrastructure Estimated Effort: Low-Medium

Option 2: Tableau MCP via Claude Desktop (Local)

Approach: Configure Tableau MCP server locally in Claude Desktop's MCP config, build artifacts that query Tableau from the desktop app. Pros: Data stays local, works with on-prem Tableau Server, full control Cons: Not shareable, only works on machines with Claude Desktop configured, may not support Live Artifacts refresh Estimated Effort: Low

Option 3: Hybrid — Tableau MCP + Google Sheets Bridge

Approach: Use Tableau MCP for complex queries/exploration, but pipe key metrics to Google Sheets (which we already have MCP for), then build Live Artifacts from Sheets as a simpler data source. Pros: Sheets is already integrated, simpler auth, team can edit/annotate, works as backup if Tableau MCP has issues Cons: Extra step, data staleness if Sheets aren't updated frequently, duplicates some Tableau functionality Estimated Effort: Medium

Chosen Approach

Decision: TBD — Need Cole's input on:

1. Are we on Tableau Server (on-prem) or Tableau Cloud?
2. Which dashboards/data sources would benefit most from Live Artifacts?
3. Security stance — comfortable with data flowing through Claude's web infrastructure?

Implementation Plan (Draft)

Phase 1: Setup & Auth

• ☐ Determine Tableau Server vs Cloud setup
• ☐ Generate Personal Access Token (PAT) for MCP access
• ☐ Install/configure Tableau MCP server (local first for testing)
• ☐ Verify connection — list available workbooks and data sources

Phase 2: Prototype Live Artifact

• ☐ Pick one high-value dashboard (e.g., daily sales summary or Meta Ads performance)
• ☐ Build a Live Artifact in Cowork that queries Tableau data
• ☐ Test auto-refresh behavior
• ☐ Evaluate data accuracy vs native Tableau view

Phase 3: Expand & Operationalize

• ☐ Identify 3-5 reports that benefit from Live Artifacts
• ☐ Set up team access if on Team/Enterprise plan
• ☐ Document which artifacts exist and what they show
• ☐ Consider scheduled bot that checks artifact health

Acceptance Criteria

Must Have:

• ☐ Tableau MCP connected and querying live data
• ☐ At least one Live Artifact dashboard working in Cowork
• ☐ Data matches what Tableau shows natively

Should Have:

• ☐ 3+ artifacts covering key business metrics (sales, ads, inventory)
• ☐ Documentation for team on how to use/create artifacts

Nice to Have:

• ☐ Automated alerts when artifact data shows anomalies
• ☐ Integration with existing morning routine scripts

References

• Tableau MCP official repo: https://github.com/tableau/tableau-mcp
• Getting started guide: https://tableau.github.io/tableau-mcp/docs/getting-started
• Claude Cowork connectors: https://support.claude.com/en/articles/11176164-use-connectors-to-extend-claude-s-capabilities
• Custom MCP connectors: https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp
• InterWorks practical guide: https://interworks.com/blog/2025/10/28/connecting-llms-to-tableau-a-practical-guide-for-using-tableau-mcp/

Execution Log

2026-04-20 (Initial Research)

• Confirmed Live Artifacts launched today via @claudeai tweet
• Found official Tableau MCP server (github.com/tableau/tableau-mcp)
• Key requirement: Personal Access Token from Tableau Server/Cloud
• Created this plan, awaiting Cole's input on Tableau infrastructure details