source
Airbyte
ELT platform with 600+ connectors — source-available under Elastic License 2.0 (not a standard OSI license). Implement the Source extension-point contract to feed anything Airbyte ingests straight into a pipeline.
elt cdc
← Back to top Community-buildable Community-buildable Community-buildable Community-buildable Community-buildable Community-buildable Shipped Community-buildable Community-buildable Community-buildable
Connectors
System Connectors
The external systems data enters and leaves through — and the provin side that receives them.
Connections are implemented against four extension-point contracts: Source / Sink / Transport / Observer.
"Shipped" means the implementation exists on the current OSS main; "community-buildable" marks systems connectable by implementing an extension-point contract.
The four extension points
Each point maps to a contract in the OSS pipeline/ tree — Process, the sink Writer, the transport Publisher / Subscriber, and contract.ProcessObserver — and the shipped components double as reference implementations.
Source
Where external data enters a pipeline.
Shipped: ingest / aggregate
Sink
Where results leave. In production / archival kinds, only verified verdicts pass (fail-closed).
Shipped: console / file
Transport
Record transport between processes.
Shipped: NATS (Core NATS)
Observer
Subscription hook for processing events — fire-and-forget, best-effort telemetry. Audit-critical persistence stays on the synchronous store paths.
Shipped: logobserver
source
Meltano
OSS ELT built on the Singer standard. The CLI / GitOps choice with 600+ taps. Receive Meltano output at the Source extension point to feed pipelines.
elt singer oss
source
Debezium
OSS framework for capturing change streams from PostgreSQL, MySQL, MongoDB, and others. Subscribe to Debezium events at the Source extension point to bring DB changes into a pipeline in near real time.
cdc oss
source sink
Webhook
A general-purpose connection shape that turns any HTTP push into an entry or exit. Receive SaaS event notifications and external triggers at the Source extension point, and deliver pipeline output to arbitrary HTTP endpoints as a Sink.
http push
source sink
Kafka / Redpanda
The staple streaming broker stack. Subscribe to existing topics at the Source extension point and write results to other topics through a Sink. Drop provin on top of an existing Kafka / Redpanda cluster — keep your broker choice and gain the provenance layer.
streaming broker
source sink
Apache Pulsar
OSS streaming platform with strong multi-tenant and geo-replication support. Connect at both the Source and Sink extension points — a good fit for geo-distributed pipeline deployments.
streaming broker oss multi-tenant
transport
NATS
Lightweight, easy-to-self-host OSS PubSub. Core NATS ships as the default
Transport in provin OSS (pipeline/transport/nats), carrying records between
Pipeline Processes — the internal transport seam, not an external data
connector, and swappable with SQS/SNS. It is part of the quickstart stack.
Exactly-once / durable delivery via JetStream is noted in the source as a follow-up — what ships today is Core NATS.
streaming broker oss
source sink
Bento
warpstreamlabs.github.io/bento/ ↗
The MIT fork of Benthos. Single-binary stream processor with declarative Bloblang mappings and 200+ processors. Connect it as upstream shaping on the Source side, or downstream delivery on the Sink side.
processing oss
sink
Apache Iceberg
The de facto OSS lakehouse table format. Writing pipeline output to Iceberg tables is implementable against the Sink extension-point contract.
lakehouse table-format oss
sink
MCP (Model Context Protocol)
An open protocol for exposing structured data to AI agents. Surface a pipeline’s output from the Sink extension point as an MCP server, so LLM agents consume provenance-backed data.
ai-agent oss