| name | tfl-generator |
| description | Generate Tableau Prep flows (.tfl/.tflx) using cwprep MCP Server or Python SDK. |
TFL Generator
MCP Server (Recommended)
When the cwprep MCP Server is connected, use MCP Tools directly:
| Tool | Purpose |
|---|
generate_tfl | Generate .tfl/.tflx file from a flow definition |
validate_flow_definition | Validate definition before generating |
list_supported_operations | List all supported node types and parameters |
Before generating, read MCP Resources for context:
| Resource | Content |
|---|
cwprep://docs/api-reference | SDK API reference (database + file connections, all transforms) |
cwprep://docs/calculation-syntax | Tableau Prep formula syntax (differs from SQL!) |
cwprep://docs/best-practices | Common pitfalls and flow design rules |
Supported Connection Types
- Database: MySQL, PostgreSQL, SQL Server (SSPI + username/password)
- File: Excel (.xlsx/.xls), CSV — auto-detected from extension
Output Formats
.tfl — Standard flow (references external data).tflx — Packaged flow (embeds data files inside)
Fallback: Python SDK
If MCP is unavailable, use the SDK directly:
from cwprep import TFLBuilder, TFLPackager
builder = TFLBuilder(flow_name="My Flow")
conn_id = builder.add_connection(host="localhost", username="root", dbname="mydb")
orders = builder.add_input_table("orders", "orders", conn_id)
file_conn = builder.add_file_connection("returns.xlsx")
returns = builder.add_input_excel("Returns", "Sheet1", file_conn)
joined = builder.add_join("Join", orders, returns,
["product_id", "order_id"], ["product_id", "order_id"])
builder.add_output_server("Output", joined, "My_Datasource")
flow, display, meta = builder.build()
TFLPackager.save_to_folder("./output/my_flow", flow, display, meta)
TFLPackager.pack_zip("./output/my_flow", "./output/my_flow.tfl")
See the project README for full SDK documentation.
