# DuckLake

> MIT lakehouse table format on SQL + Parquet — a DuckDB extension

**Category:** Databases  
**Pricing:** open-source  
**URL:** https://ducklake.select  
**GitHub:** https://github.com/duckdb/ducklake  
**Added:** 2026-10-08

---

## The Hook

DuckLake is an open lakehouse format from the DuckDB team: table metadata lives in an ordinary SQL database, data lives in Parquet files on disk or object storage — snapshots, time travel, and schema evolution without a custom catalog server.

## What's Good

MIT (Stichting DuckDB Foundation). DuckLake v1.0 is out; `INSTALL ducklake;` then `ATTACH 'ducklake:metadata.ducklake' AS lake (DATA_PATH 'files/');` and use plain SQL. Catalog can be DuckDB, PostgreSQL, SQLite, or MySQL; Parquet on local disk or S3-compatible storage; multiple clients against one catalog. The first lakehouse table format on Stack Gems. ★~3k.

## The Catch

Needs a separate catalog database you run and back up (DuckDB file for one writer, Postgres/MySQL for multi-client). Engines outside DuckDB support it far less widely than Iceberg or Delta Lake — check your Spark/Trino/warehouse story before committing. Younger format with a shorter production track record. Vs the duckdb gem: DuckDB is the engine; DuckLake is the table format and catalog layer on top.

## Verdict

MIT SQL-catalog lakehouse format when DuckDB is your engine and Iceberg’s catalog stack feels heavy — confirm multi-engine support first.