WrenAI
GenBI (Generative BI) for AI agents, an open-source, governed text-to-SQL through an open context layer that turns natural-language questions into trusted dashboards, charts, and SQL across 20+ data sources, such as BigQuery, Snowflake, PostgreSQL, ClickHouse, Amazon Redshift, Databricks and more.
WrenAI is an open-source context layer that lets AI agents write accurate, governed SQL across 20+ data sources by modeling tables, relationships, and rules once in MDL.
WrenAI is an open source project that acts as a context layer between AI agents and your business databases. The core problem it solves is that AI models like Claude or ChatGPT struggle to write accurate SQL (the language for querying databases) against real business data because they do not understand what the tables and columns actually mean — overlapping naming, inconsistent definitions, and complex relationships confuse them. WrenAI lets you model your business data once using MDL (Modeling Definition Language) — defining entities, relationships, calculations, and access rules — and then any AI agent queries through that shared understanding rather than trying to interpret the raw database directly.
The system is powered by a Rust engine built on Apache DataFusion (a query execution engine) that translates modeled SQL queries and runs them against over 20 data sources including PostgreSQL, BigQuery, Snowflake, MySQL, DuckDB, Databricks, Redshift, ClickHouse, Athena, Spark, and others. It can be used as a Python SDK, a command-line tool, or as a WebAssembly (WASM) module that runs in the browser. There are also integrations for LangChain (a popular AI framework) with more framework integrations listed as coming soon.
You can install WrenAI's agent skills using the npx skills add command and let an AI coding assistant like Claude Code or Cursor handle the onboarding setup. The core components are licensed under Apache 2.0, built by Canner.
Where it fits
- Give an AI agent reliable text-to-SQL access to a production warehouse.
- Model business metrics once in MDL and reuse them across Claude, Cursor, and ChatGPT.
- Build a GenBI app that turns natural language into charts and dashboards.
- Embed the WASM engine in a browser tool that runs governed SQL queries.