Run any transformation, stateful or stateless, simple or complex, battle-tested at enterprise scale.
Fully flexible transformations
Run stateful and stateless transformations with long time windows.
Run in production at scale
GlassFlow is built to run on TBs of production data every day.
The performance metrics prove it.
With a single pipeline (44 TBs daily)
Linear Scaling - with 51k rps/replica
Latency for 80% of transformations
Resource costs per TB
(c4d-standard-16)
The only purpose-built open source streaming ETL that moves and transforms your data.
Dedupe
Fast joins
Stateless transformations
Late event handling
Reduces load on ClickHouse
Low maintanance effort
Dead-letter-queue
Pipeline
Observability
Open source
Enterprise Support
Deployment Service
CH Kafka Table Engine
Needs RMT
ClickPipes
for Kafka
Needs RMT
Self-Built Go Service
Needs Custom Code
Needs Custom Code
Needs Custom Code
Needs Custom Code
Vector.dev
Designed specifically for ClickHouse. With GlassFlow you fix data correctness for TBs of data before ClickHouse sees the data.
Managed Integrations
Connect GlassFlow to integrations like, Kafka, OTEL, Postgres, and more. Our ClickHouse connector uses the native protocol for the best performance and experience.
⮕ See all integrations
Streaming Transformations
A lightweight built-in state store enables low-latency, in-memory dedupe and joins with time-windowed context retention. GlassFlow also includes standard stateless transformations via the EXPR library.
Dead-Letter-Queue keeps your pipeline running
Isolate faulty events without interrupting data flow, and re-run them effortlessly after fixes.
Enterprise Workloads
Ingest TBs of data into ClickHouse while transforming. GlassFlow runs on Kubernetes with a Dead Letter Queue and built-in pipeline observability.
Cost Efficient Infra Footprint
A lightweight platform with no clusters to manage, delivering high performance at low operational cost.
Use Case Spotlight:
Build your own end-to-end OSS observability stack
An open source stack that cuts your observability costs to near zero. Try out our tutorial and self-host the highly scalable and open source observability stack.
Feel free to contact us if you have any questions after reviewing our FAQs.
