duckle

Why we included this project

Most ETL tools push you into a vendor cloud or per-row pricing. Duckle runs on your own servers or cloud account instead, and compiles every pipeline to SQL on DuckDB. You can author a flow on a visual canvas, in Python, or in SQL, then ship that same file to a headless runner that executes on a schedule in Docker or on a box you own, with a web console, roles, and an audit trail. It carries the features production teams usually have to stitch together separately: roughly 380 connectors, dbt support, change data capture, data quality checks, reverse ETL, and lineage. An MCP endpoint lets AI agents call your pipelines or read the data underneath them. Because each pipeline is one file in git, the work outlives whoever wrote it, which gives small teams portability without platform lock-in.