Build verified dashboards from natural language questions using an open-source engine
Imagine pulling charts directly from database numbers to share with your team. Unlike existing AI tools that simply generate SQL statements, this tool understands business rules and deploys verified results as dashboards. It processes the entire flow in one go: accepting natural language questions, writing SQL based on defined metrics, and generating a link that opens in a web browser.
Business meaning beyond the schema
A database schema only tells you where the data is located, not what it means in a business context. This tool manages approved metric definitions, enumerated values, units, and approved join conditions in YAML and Markdown files. This knowledge is stored in a Git repository, allowing it to be reviewed and version-controlled like code. As a result, the term 'revenue' is interpreted with the same meaning by all agents and team members.
The flow from question to deployment
Users ask questions in natural language, and the agent loads the project's business context to generate SQL. The generated SQL undergoes a validation process before execution, providing structured hints if errors occur. In the final step, the results are converted into a dashboard that runs in the browser and deployed to Vercel or Cloudflare Pages. This process is controlled by commands like wren query and wren skills get genbi, so users do not need to learn a separate BI tool UI.
Integration with existing stacks
This tool operates as a layer on top of AI agent frameworks you are already using, such as Claude Code, Cursor, and LangChain. It supports more than 22 data sources, including BigQuery, Snowflake, PostgreSQL, and ClickHouse, through the Apache DataFusion engine. You can connect and use it in your current environment without modifying existing data warehouses or transformation pipelines. Additionally, it improves answer accuracy by searching past queries and examples using a locally running LanceDB memory index.
Team deployment and security boundaries
Authentication tokens use one-time tokens that expire after 600 seconds, ensuring no permanent credentials are left on disk. The core engine and SDK are free under the Apache-2.0 license, but row/column-level security, GenBI UI, and advanced audit logs are available in the commercial versions, Wren AI Cloud or Enterprise Plus.
Considerations before adoption
This tool is suitable for teams that want to increase the reliability of agent answers in data warehouses with complex business logic. Conversely, it is not suitable for creating one-off charts from a single CSV file or for environments that allow agents to guess SQL without governance. Before adoption, you must verify whether your data source is included in the list of 22 supported sources and clearly distinguish whether the required security features are in the open-source or commercial version.
