Description
Let an AI query your company database and it guesses at table names and computes metrics wrong, leaving reports no one dares trust for the boss? WrenAI adds a layer between data and AI that actually understands the business—an open-source generative BI engine that turns natural-language questions into governed, trustworthy SQL, then all the way into charts and dashboards. You ask in plain language; it queries correctly and visualizes it.
The key is its semantic and context layers: how each metric is computed, what fields mean, and which business conventions apply — all written as reviewable YAML in a Git repo you own, so the AI answers from those definitions instead of guessing at schemas. It supports 20+ warehouses and can serve as a capability layer for agents like Claude Code, Cursor and MCP. Worth adopting for data teams who want both people and AI to self-serve queries without losing control. Apache 2.0, open source and self-hostable.
Natural-language questions: ask in plain language and it generates SQL, runs it, and assembles the result into charts and shareable dashboards.
Semantic & context layer: approved metric definitions, enums, units, joins and the tribal knowledge buried in docs and chat, all as reviewable YAML so the AI knows what the data means.
Governance & guardrails: definitions live in a Git repo you own, every change is reviewable, and AI-generated SQL stays inside your rules.
20+ data sources: connects to BigQuery, Snowflake, PostgreSQL, ClickHouse, Redshift, Databricks and other major warehouses.
Feed your AI agents: give Claude Code, Cursor, MCP clients and LangChain a trusted data-access layer instead of guessing at schemas.
The key is its semantic and context layers: how each metric is computed, what fields mean, and which business conventions apply — all written as reviewable YAML in a Git repo you own, so the AI answers from those definitions instead of guessing at schemas. It supports 20+ warehouses and can serve as a capability layer for agents like Claude Code, Cursor and MCP. Worth adopting for data teams who want both people and AI to self-serve queries without losing control. Apache 2.0, open source and self-hostable.
Features
Natural-language questions: ask in plain language and it generates SQL, runs it, and assembles the result into charts and shareable dashboards.
Semantic & context layer: approved metric definitions, enums, units, joins and the tribal knowledge buried in docs and chat, all as reviewable YAML so the AI knows what the data means.
Governance & guardrails: definitions live in a Git repo you own, every change is reviewable, and AI-generated SQL stays inside your rules.
20+ data sources: connects to BigQuery, Snowflake, PostgreSQL, ClickHouse, Redshift, Databricks and other major warehouses.
Feed your AI agents: give Claude Code, Cursor, MCP clients and LangChain a trusted data-access layer instead of guessing at schemas.
