Cwprep

What is Cwprep?

Cwprep is a Python-based engine that enables text-to-PrepFlow generation by reverse-engineering the .tfl JSON structure. It provides a built-in MCP server, acting as a bridge between LLMs (such as Claude and Gemini) and Tableau Prep, allowing users to generate, modify, and build data cleaning flows through natural language conversations or Python scripts without opening the GUI.

How to use Cwprep?

Install with pip install cwprep or use uvx (recommended) which auto-downloads the MCP extra. Start the MCP server locally with cwprep-mcp for stdio transport, or with cwprep-mcp --transport streamable-http --port 8000 for remote HTTP access. Configure supported clients (Claude Desktop, Cursor, VS Code, Windsurf, Gemini CLI, etc.) to point to this server using the uvx method.

Key features of Cwprep

• Generates TFL/TFLX files from natural language flow definitions
• Translates TFL flows to equivalent ANSI SQL
• Validates flow definitions before file generation
• Connects to MySQL, PostgreSQL, SQL Server, Excel, and CSV
• Supports joins, unions, filters, calculations, pivots, and aggregations
• Packages flows as .tflx archives with embedded data files

Use cases of Cwprep

• Build data cleaning pipelines via conversational AI in Claude Desktop
• Automate Tableau Prep flow generation from Python scripts
• Translate existing TFL flows into SQL for review or validation
• Design and modify multi-step data flows without manually opening Tableau Prep

FAQ from Cwprep

What does Cwprep do that alternatives don't?

Cwprep enables generation of Tableau Prep Flow (.tfl) files directly from natural language or Python code, acting as a bridge between LLMs and Tableau Prep without requiring the GUI.

What are the runtime requirements for Cwprep?

Python 3.8 or higher is required. The MCP server extra can be installed via pip install cwprep[mcp] or auto-downloaded using uvx.

What transports and authentication does the MCP server support?

The server supports stdio transport (local) and Streamable HTTP transport (--transport streamable-http --port 8000). Authentication is not detailed in the README.

Where are generated flow files saved?

By default, the SDK and MCP output only the final .tfl or .tflx archive. Use save_to_folder() explicitly to obtain the exploded folder for inspection.

Can Cwprep connect to live databases?

Yes, it supports database connections to MySQL, PostgreSQL, and SQL Server via add_connection(), with default settings configurable in config.yaml.