Adapt Kafka to new consumer requirements automatically.

Your Kafka partitioning strategies become harder to change as data accumulates. kplex lets you reorder, repartition, and join your Kafka topics on-the-fly. This allows you to unlock concurrency and perform correct stateful processing across partitions.

Virtual Partitions

Repartition topics without additional storage or latency costs.

Cross-partition Read

Records provided in event time order, even across topics and partitions.

Topic-Topic Joins

Streaming topic-topic joins on keys or on payload attributes.

Push-based Access

Multi-threaded polling of input topics at configurable intervals.

Read our introductory blogpost

Use cases

Increase consumer concurrency.

Kafka encourages you to choose a reasonable number of partitions for your topics. But computers are fast and more partitions can unlock more concurrency.

Use kplex to repartition a topic with a small number of physical partitions into many virtual ones for highly data-parallel tasks. Without touching your consumer logic.

Pull related records consistently.

Stateful processing is easier when everything is in a single topic, especially when order matters. But normalization allows for scale, simpler schemas, and granular retention policies.

kplex can present multiple normalized input topics as a single, denormalized virtual topic, without you having to worry about consistent streaming joins.

Decouple physical from logical partitioning.

Uniformly partitioned records prevent hotspots and thus ensure smooth operations. On the other hand, partitioning by domain attributes is crucial for consistent, stateful processing.

Use kplex to repartition topics on the fly, without duplicating data and without intermediate partitions. You will receive records in event time order, for consistent aggregates.


499 / month
  • Virtual Partitions
  • Cross-partition reads
  • Topic-Topic Joins
999 / month
  • Virtual Partitions
  • Cross-partition reads
  • Topic-Topic Joins
  • Run across multiple machines
7999 / week

Event sourcing and consistent stream processing have their share of challenges.

We help you model your data, improve performance, and verify correctness.