Connectors
A connector is a pre-built integration to an external system that a Pipeline reads from or writes to. Connectors handle authentication, batching, and retry behavior for a specific system so pipeline authors don't have to.
Connector direction
- Source — emits events into a pipeline (e.g. a change-data-capture stream from Postgres).
- Destination — receives events from a pipeline (e.g. writing rows into Snowflake).
- Bidirectional — a small set of connectors, like the generic Webhook connector, can act as either.
Built-in connector catalog
| Connector | Direction | Auth |
|---|---|---|
| Postgres | Source | Connection string + replication slot |
| Amazon S3 | Source / Destination | IAM role or access key |
| Snowflake | Destination | Key-pair auth |
| Kafka | Source / Destination | SASL/SSL |
| Webhook | Source / Destination | Signed request or bearer token |
Rotating a connector's credentials pauses any pipeline using it for the duration of the rotation. Plan credential rotations outside peak processing windows.
Connector health
Each connector reports a health status independent of any single pipeline: healthy, degraded, or down. A connector shared across multiple pipelines only needs to be fixed once — every pipeline referencing it recovers automatically.
See also
Connectors are configured per pipeline in the Connecting a data source guide, and referenced directly in the Connectors endpoint of the API reference.