wren-engine

MCP.Pizza Chef: Canner

Instead of letting a model invent its own database queries against raw tables, this sits in between and holds the meaning: what counts as an active customer, how net revenue is worked out, which tables relate to which. It reaches Postgres, BigQuery, Snowflake, ClickHouse, S3 and roughly a dozen more sources, and it ships a companion server with a read-only switch so an assistant can look but not change anything. Note that this repository is now archived; the work moved into the Canner WrenAI project.

Archived · repository archived; work moved into Canner WrenAI

Data

Use This MCP server To

Ask for net revenue and get our official definition, not a guess Query the warehouse without knowing which tables to join Let my assistant answer data questions inside Claude Code Keep an assistant read-only while it explores our numbers Give the same business definitions to every tool we use

README

Wren Engine — Archived

This repository has been merged into Canner/WrenAI under the core/ directory and is now archived (read-only).

New issues and PRs → Canner/WrenAI Full migration details (path mapping, archived modules, rationale) → discussion #1592

Wren Engine

The open context engine for AI agents

Wren Engine is the open foundation behind Wren AI: a semantic, governed, agent-ready context layer for business data.

wren-engine FAQ

Is this repository still active?

No, it is archived and read-only. Development continues inside the Canner WrenAI project under its core directory.

Which apps does this work in?

It is used from Claude Code, Claude Desktop, Cursor, VS Code, and Cline, among other MCP-capable tools.

Which databases can it reach?

Postgres, MySQL, BigQuery, Snowflake, Redshift, Databricks, ClickHouse, DuckDB, Trino, Oracle, SQL Server, S3 and more.

Can I use this to ask about company numbers safely?

Yes. A read-only mode blocks anything that would change your setup or reveal the raw table details underneath.

Do I need a paid key?

No key is required for the open source engine, though you do supply the login details for whichever database you connect.

How hard is the setup?

This is a developer job. You run the engine yourself, describe your data in its model file, then connect an assistant to it.

Is this something my assistant connects to, or something that connects to my assistant?

It ships an MCP server of its own, so your assistant connects to it the same way it connects to any other tool.