← Back to Discover

Wren AI

An open-source context engine that sits between AI agents and your data, turning questions into governed SQL using shared definitions. Runs through a CLI and MCP server; paid cloud and enterprise plans add a UI.

Data analysisMCP servers
Visit website ↗

What it does

Wren AI is an open-core project built around a context engine. The engine sits between your AI agents and your data and applies shared definitions, kept as MDL files in git, at query time. A question arrives, the definitions are applied, and governed SQL runs against your data where it already lives. The site says data is queried in place and never moved or copied. The open-source part is aimed at individual developers and is used through a command-line tool and an MCP server, with no UI. It includes context modelling, a text-to-SQL pipeline and more than 20 connectors, and it works with any language model you bring. The site names Snowflake, BigQuery, PostgreSQL and Databricks among the sources, and Claude and Cursor among the agents that connect over MCP. Commercial plans run the same engine and add what the open-source part leaves out: a GenBI interface with apps, an agentic mode and dashboards, embedded analytics and APIs, user and group access control, row- and column-level security, audit and support. They are offered as Wren AI Cloud, with a 7-day free trial, or as a self-hosted Enterprise Plus deployment inside your own network.

Capabilities

  • Open-source context engine with MDL definitions that live in git
  • Text-to-SQL pipeline that runs governed queries against your own data
  • MCP server and CLI so agents and scripts can query through the engine
  • 20+ data source connectors, queried in place
  • Works with any language model you choose
  • Cloud or self-hosted commercial plans add GenBI apps, dashboards and security controls

When to use it

  • You are building your own agent and want it to query a warehouse with shared metric definitions
  • You want to try the engine locally for free before choosing a paid platform
  • You need access control and row-level security on AI-driven analytics inside your own network