Stateful stream processing for Kafka, built to scale up — not out.
Cut the distribution tax: fewer moving parts, simpler operations, more throughput.

StoatFlow is a Java stream processing library with the Kafka Streams DSL, deployed as a single replica. It implements the full DSL —
KStream, KTable, joins, windowed aggregations, the Processor API — so existing topologies port over with a dependency swap and a config
cleanup, not a rewrite. On top you get StoatFlow extensions: Flink-style watermarks and timers, scheduled sources, and side outputs.

Built on modern Java 25 — virtual threads, structured concurrency, FFM. Lane parallelism is decoupled from Kafka partition count, state is
global (readable from any virtual thread), repartitioning happens in-memory with no extra Kafka topics, and exactly-once commits flow
through one coordinator. The result: no rebalancing, no state migration, no repartition round-trips — fewer moving parts.

Kafka Streams and Apache Flink are excellent at what they do — Kafka Streams as a lightweight client library, Flink at massive scale.
StoatFlow targets the wide middle: Kafka-native workloads that fit on one modern machine and benefit more from operational simplicity than
cluster elasticity.

The numbers, on 8 vCPU: a 5-table foreign-key-join chain runs at 11–13.6× lower P95 latency, 4.4–7.8× less container memory, and 1.5–3.4×
less CPU than Kafka Streams — at throughput parity. State restoration is 1.45–1.65× faster on the same hardware.

Built by a three-time Current speaker — 18 years in distributed systems, deep in Kafka Streams, and author of the open KSTD topology-design
standard.

Commercial, JVM-native, single-instance. See the benchmarks and docs at stoatflow.io.

StoatFlow

Org chart

Teams

Offices

StoatFlow

Org chart

Teams

Offices

Related companies