WrenAI

What it does

WrenAI sits between a person and the warehouse: it maps business terms to tables, generates SQL, and shows the mapping so an analyst can check what it did. It is open source and can run inside your own network.

The value is the semantic layer, not the model. If nobody defines what "active customer" means, the generated SQL will be syntactically neat and commercially wrong.

Standout features

• Business-term to schema mapping
• Generated SQL with visible reasoning
• Self-hosted open-source deployment
• Read-only by design

Main job: answer a plain-language question about your own data with SQL you can inspect.

Best for

Use it when analysts spend their day writing the same joins for ad-hoc questions from the sales or sourcing team.

Getting started

Connect a read-only warehouse user, define the twenty business terms people actually ask about, then test each term against a known number before opening it to the team. Keep it read-only.

Watch out for

Watch out: never give it write credentials, and inherit source permissions — a semantic layer that bypasses row-level access is a data leak. Spot-check generated totals against the finance report until the mappings settle.