Pipes in Tinybird usually expose query results as HTTPS endpoints. A sink pipe inverts that flow: the result of the pipe is written to an outside system such as object storage, a data warehouse, or a Kafka topic, on a cron-like schedule defined alongside the pipe.

This is the natural primitive for reverse ETL, scheduled exports, and feeding offline analytics or batch ML pipelines from real-time data. Because the query runs inside Tinybird, sinks inherit the same incremental processing model as the rest of the workspace.

Sink pipes differ from copy pipes in that the destination is external. They differ from endpoints in that the data is pushed, not pulled.