WrenAI

WrenAI is a generative business intelligence (GenBI) engine that gives the coding agents and agent frameworks you already use a governed semantic layer plus an AI context layer. Agents use it to turn business questions into correct SQL, execute it across more than 22 data sources, and publish the answer as a shareable dashboard. It works as a layer under those agents rather than as a separate chat interface.

Schemas tell an agent where data lives, and WrenAI adds what the data means. The semantic layer is written in the Modeling Definition Language (MDL): models, columns, relationships, views, cubes, and approved metrics. The context layer adds enums, units, approved joins, worked examples, and knowledge pulled from docs and chat threads. All of it lives as reviewable YAML and Markdown in a Git repository you own, so metric changes are reviewed as pull request diffs.

Queries are planned against the MDL and checked before execution through schema-aware retrieval, dry-plan validation, row limits, value profiling, and structured errors with hints. The engine is built on Apache DataFusion and connects to BigQuery, Snowflake, PostgreSQL, ClickHouse, Amazon Redshift, Databricks, DuckDB, and others. Dashboards are browser-side apps built on a WebAssembly build of the engine and deployed to Vercel or Cloudflare Pages.

Agents learn the workflow through a small skill stub that fetches full guides from the CLI on demand, and it works with Claude Code, Cursor, Cline, Codex, MCP clients, and LangChain. SDKs cover LangChain and LangGraph and Pydantic, with a reference Python integration for other stacks. The engine, MCP server, CLI, and connectors are open source. Row- and column-level security, the GenBI UI, embedded dashboards, and managed or air-gapped deployment are commercial, delivered as Wren AI Cloud or a self-hosted enterprise edition, and Git Sync turns the same repository into a team-wide deployment on either.