PeerDB is an open-source change data capture engine built specifically for PostgreSQL sources, acquired by ClickHouse in 2024 and kept as a distinct component powering the managed Postgres CDC in ClickPipes.
Self-hosted as a Flow API with workers, a Postgres catalog, and Temporal for orchestration; sources need wal_level=logical, a publication and slot, and primary keys. Free to self-host; managed as ClickPipes, billed on compute hours plus data ingested.
How PeerDB answers the questions Change Data Capture turns on.
| How it works | |
| Initial snapshot | Parallel initial load, partitioned by CTID |
| End-to-end lag | Near real-time; streams from the slot on a configurable sync interval |
| Delivery guarantee | Exactly-once |
| Schema drift | ADD COLUMN propagates; drops and type changes don't |
| Running it | |
| What you run | Flow API and workers, a Postgres catalog and Temporal |
| Connections | |
| Sources | PostgreSQL and MySQL; MongoDB in preview |
| Source requirements | wal_level logical, a publication and slot, and a primary key |
| Transformations | None in the pipeline; Lua row transforms when writing to queues |
| Destinations | ClickHouse first; other warehouses and queues |
| Cost | |
| Billing unit | Free self-hosted; ClickPipes bills compute hours plus GB ingested |
vs PeerDB: Open source (permissive) · Operational complexity: High · Java, Scala, Python
vs PeerDB: Commercial · Managed · Serverless · Subscription · Batch · Streaming
vs PeerDB: Open source (permissive) · Operational complexity: High · Streaming · Java