Code-First Intelligent Data Flow Infrastructure
Unlike the official Tableau Prep AI Assistant, cwprep breaks free from the black-box graphical interface limitations by providing a declarative, code-based foundation. It translates natural language intents instantly into fully auditable pipelines translated to ANSI SQL—bringing total transparency to your architecture team alongside the freedom of multi-model MCP integrations.
Core Capabilities
Beyond mere AI Generation
The core of cwprep is a deeply robust "Headless" Prep operations engine. It treats Tableau file formats as first-class citizens, ensuring programmatically valid XML tree manipulations based strictly on XSD specifications.
Extreme Reverse Engineering: Built-in
ExpressionTranslatorinstantly decompiles over 30 gnarly Tableau Prep functions (e.g. DATEPART, ZN) back into universally compatible regex configurations and native SQL.Safety Temp Backup Locks: Given the unreliability of asynchronous CI/CD systems, cwprep deploys deterministic file backup locking upon initialization to prevent destructive .tfl corruption during pipeline failures.
Declarative Configuration Logic
Embracing modern Infrastructure as Code (IaC) doctrines, cwprep calculates dynamic node paths based on terminal state requirements, completely bypassing rigid procedural mappings.
DevOps Ready Linkages
Inject Prep compilation directly into CI/CD build scripts. Bring sweeping standardization and continuous integration to fragmented BI assets.
Capability Matrix
| Feature | cwprep (Headless) | Tableau Prep AI | Advantages |
|---|---|---|---|
| Under-the-hood Transparency | Yes (100% translated to SQL) | No (Internal GUI Blackbox) | Allows data flows to be exposed as SQL for DBA architectural review, completely independent of the client. |
| Headless Batch Processing | Yes (Code & Prompt Driven) | No (Tied to Visual Interface) | Instantly build dozens of pipelines without waiting for GUI renders, boosting productivity massively. |
| Privacy & LLM Base | Yes (Attach any MCP compatible LLM) | No (Bound to official cloud) | Freely switch between Claude, GPT, or air-gapped DeepSeek models representing zero data leakage. |
Datacooper
Yes (100% translated to SQL)
Official AI
No (Internal GUI Blackbox)
Allows data flows to be exposed as SQL for DBA architectural review, completely independent of the client.
Datacooper
Yes (Code & Prompt Driven)
Official AI
No (Tied to Visual Interface)
Instantly build dozens of pipelines without waiting for GUI renders, boosting productivity massively.
Datacooper
Yes (Attach any MCP compatible LLM)
Official AI
No (Bound to official cloud)
Freely switch between Claude, GPT, or air-gapped DeepSeek models representing zero data leakage.
Product Roadmap
Phase 1: Foundation
Core node generation & full ANSI SQL translation.
Phase 2: Collaboration
Launch MCP shared server mode for team templating.
Phase 3: Governance
Integrate data quality gateways to stop skewed processes.
Pricing Options
- - Core TFL Builder Engine
- - Unlimited MCP Generation
- - Community Github Support
Frequently Asked Questions
Q: Is cwprep bound to a specific LLM?
A: No. cwprep is model-agnostic and supports Claude, ChatGPT, Gemini, and DeepSeek, as long as they support standard reasoning.
Q: Can the generated .tfl files be opened directly in Tableau Prep?
A: Yes. cwprep generates official .tfl and .tflx files that can be opened and run natively without plugins.
Q: Does it support complex row/column transformations (Pivot)?
A: Absolutely. cwprep is specifically optimized for Pivot and Unpivot node generation.
Quick Start
View Comprehensive Docs# 1) Installation
pip install cwprep
# 2) MCP Server Config Example
{
"mcpServers": {
"cwprep": {
"command": "uvx",
"args": ["--from", "cwprep[mcp]", "cwprep-mcp"]
}
}
}
# 3) Prompt Example
"Connect to intranet SQL Server (SSPI) to build a multi-database cleaning flow"
"Remove invalid columns, then execute Unpivot on month data"
"Finally output ANSI SQL with standard comments for audit review"