# Tableau MCP Locally: Why I Won't Use Claude for My Company Data

> This is content from just-tim, the data-and-analytics channel by Tim Ngwena (formerly 'Tableau Tim'). Tim has 12+ years of hands-on BI experience and covers Tableau most of all, plus Power BI, Looker, Hex, SQL and data modelling, the analytics industry, and the craft of doing the job — always tool-agnostic and honest about the trade-offs.

- **Author:** Tim Ngwena (just-tim, https://just-tim.com/about)
- **Published:** 2025-10-28
- **Format:** Video · 23 min watch · transcript available
- **Topics:** AI & ML, Tool strategy, Data visualisation
- **Tools:** Tableau (cloud, mcp, pulse)
- **Canonical:** https://just-tim.com/posts/tableau-mcp-locally-why-i-wont-use-claude-for-my-company-data
- **Watch:** https://www.youtube.com/watch?v=RCS4wjOR6fw

I show how to run the Tableau MCP locally on your laptop using LM Studio, rather than piping your company data through Claude or ChatGPT. I walk through choosing local models, configuring the MCP, setting the context window, and querying Tableau data sources and Pulse metrics entirely offline.

## Key takeaways

- Running the Tableau MCP through Claude or ChatGPT risks exposing your company IP and infrastructure context, since you can't verify how those services handle your data despite their service agreements.
- MCPs are model-agnostic, so you can run them locally with tools like Ollama (command-line, server-oriented) or LM Studio (a friendlier chat-style application).
- On Apple Silicon, choose MLX-optimised models and check the download panel for 'full GPU offload possible' to avoid CPU fallback and major performance drops.
- You must manually increase the context window in LM Studio for the MCP to work, balancing it against your available RAM, and each run is non-deterministic even with identical settings.
- Larger 'thinking' models like GPT-OSS 20B give faster, more capable responses and can self-troubleshoot errors, but demand more RAM than smaller models like Gemma 3 12B.

## Chapters

- 0:01 Why I won't use Claude for company data
- 1:52 Ways to run local models: Ollama vs LM Studio
- 2:42 Touring the LM Studio interface
- 5:21 Browsing and choosing models
- 7:05 Configuring the Tableau MCP
- 8:30 Loading a model and setting context window
- 12:16 First query with Gemma 3 12B
- 16:39 Switching to the OpenAI 20B model mid-chat
- 20:29 Closing thoughts and what's next

Watch the full video, read the transcript and use chapter deep-links on the page: https://just-tim.com/posts/tableau-mcp-locally-why-i-wont-use-claude-for-my-company-data

---
just-tim — Data and analytics, with a point of view. · https://www.youtube.com/channel/UC7HYxRWmaNlJux-X7rNLZyw · https://twitter.com/TableauTim · https://www.linkedin.com/in/timngwena