maybe Say goodbye to endless dragging and dropping or a new way! Watch how AI generates a complete Tableau dashboard using natural language. For the longest time, we've been accustomed to the tedious process of manually dragging fields and aligning formats in Tableau. Today, I want to share a completely new paradigm for BI development.
I highly recommend watching this demo video. You can clearly see the magic of automated dashboard construction: Natural Language Input: Inside an AI client (like Cursor or Claude), I simply typed out a plain-text requirement. AI Auto-Execution (The MCP Magic): Immediately, the AI begins to think autonomously and calls the backend tools. It flawlessly executes create_workbook (loads the template), add_calculated_field, configure_chart, and finally add_dashboard to assemble everything. The Reveal: In just a few seconds, a standard .twb file is generated. When I open it in Tableau Desktop, all data connections, chart configurations, and dashboard layouts are perfectly rendered. Zero manual drag-and-drop.
There is a core technical detail hidden here, which is the secret behind the perfect layout in the video: Anyone familiar with Tableau knows that its underlying XML layout logic is extremely complex and deeply nested, making it incredibly difficult for an AI to write accurately.
To achieve the precise dashboard layout shown in the demo, I specifically developed a custom "Layout-to-JSON" parser.
It cleverly abstracts Tableau's tedious container nesting into a declarative JSON structure, much like frontend FlexBox. This means the AI only needs to output a clearly structured JSON dictionary, and my parser instantly translates it into a flawless Tableau dashboard layout! Behind all of this is my latest open-source project: cwtwb — an MCP (Model Context Protocol) Server built specifically for Tableau. It is not just a Python library; it gives your AI assistant "hands." By integrating the MCP protocol (with zero configuration, just run uvx cwtwb), your AI can now directly understand your analytical intent and build everything through conversation—whether it's rapid prototyping or complex enterprise-level layouts.
If you are interested in automated AI + BI workflows, or want to equip your Claude/Cursor with Tableau development skills, feel free to explore the source code on my GitHub.
Try running the Demo! If you think this "AI writing Tableau" concept is cool, I would appreciate your Star and feedback!
#Tableau #MCP #ModelContextProtocol #AI #Cursor #Claude #BusinessIntelligence #OpenSource #DataVisualization #DataAnalytics #datafam #gemini #google #microsoft #bi #powerbi #opentowork #cwtwb
Matthew Miller Elif Tutuk Allan Folting Roy Raviv Adam Mico Hi , I built an open-source Python SDK and mcp that allows us to generate Tableau dashboards programmatically. This is a demo now. Would love your thoughts!