Continuous replication and fan-out

One source reader. Every destination advances independently.

Committed changes enter a shared durable spool once. Each sink owns a cursor, apply strategy, error policy, and lag measurement.

Once
Decode the source

Capture cost does not multiply with sinks.

N cursors
Independent progress

Slow sinks do not redefine the others.

Durable
Backlog and catch-up

Unavailable targets recover from spool state.

Independent consumers

Isolation without duplicating capture.

Relational, event-stream, and Fabric destinations read the same ordered source history through destination-specific workers.

One-to-many delivery from one spool
Per-destination checkpoints and lag
Parallel keyed apply where safe
Source-only monitoring mode

Backpressure with a boundary

Choose what happens when retention fills.

Segments, checksums, compression, indexes, and torn-tail recovery make backlog a durable operating state with an explicit retention policy.

Configurable retention and limits
Backlog bytes and drain ETA
Independent cursor garbage collection
Pause and resume with checkpoints

Keep exploring

See the connected parts of the platform.

Map every consumer before adding another reader.

Tell us the source, destinations, latency, and outages. We will size cursors, retention, and apply.