Enterprise CDC · Self-hosted

Move terabytes.
Sync at the speed of light.

Supermetal is a self-hosted, turnkey change data capture platform. A single Rust binary that replaces the Debezium + Kafka + Flink stack and SaaS ELT tools like Fivetran and Airbyte.

Try SupermetalTalk to us

1,000 hours of free sync. No credit card. Runs on your infrastructure.

In production

Goldsky
Web3 data infrastructure
Automotive e-commerce unicorn
Multi-source lakehouse
Global automotive analytics firm
Multi-TB snapshots
Fortune 50 EV manufacturer
Multi-source CDC at scale

Product

Change data capture from every transactional source.

Snapshot, CDC, and schema evolution. Captured from your OLTP databases, delivered to warehouses, Iceberg lakehouses, or Kafka.

Sources
MongoDBMySQLOraclePostgresSQL Server
Supermetal
Single Rust binary
./supermetal
Targets
ClickHouseDatabricksDoris/VeloDBDuckDB/MotherDuckIcebergKafkaPostgresSnowflake

Engineering

Architecture that scales with your compute.

01
Rust + Apache Arrow

Columnar in-memory pipeline. No JVM, no garbage collection pauses.

02
Object storage native

Writes Parquet to S3, GCS, or Azure Blob. Targets pull via native table functions.

03
Schema fidelity

Preserves precision, scale, and complex types. No flattening to JSON or CSV.

04
Transactional consistency

Multi-table transactions stay atomic. Consumers never see partial state.

05
Parallel snapshots

Chunked across CPU cores and tables. Multi-TB initial loads in minutes.

06
Runs in your infrastructure

Kubernetes, ECS, VMs, or bare metal. Runs in your VPC. Zero egress.

Try it

Single binary. Get started in minutes.

1,000 hours of free sync. No credit card. Runs on your infrastructure.

$curl -fsSL https://trial.supermetal.io/install.sh | sh
./supermetal — trial

FAQ

Common questions.

Something else on your mind? Talk to us.

What does Supermetal replace?
Multi-system CDC pipelines built on Debezium, Kafka, and Flink, and SaaS ELT tools like Fivetran and Airbyte. One Rust binary covers both real-time CDC and batch loads.
Which use cases does it solve?
Real-time analytics, database migration, operational replication, Iceberg lakehouse ingestion, and Kafka-based CDC for existing consumers. Batch and streaming, from a single binary.
Which deployment options do you support?
The trial agent runs on any local machine, up to 1000 hours of sync. Production runs self-hosted in your cloud account or on-prem datacenter, with templates for Kubernetes, ECS, Docker, and bare metal. Contact us for air-gapped deployments.
Why self-hosted?
Self-hosted, but turnkey. A single binary that runs in your cloud, on-prem datacenter, or air-gapped network. Your data and credentials never leave your infrastructure. No telemetry, no egress fees. Most ELT vendors gate self-hosted behind enterprise contracts and months of deployment. We don't.
How does it compare to other ELT tools?
Built ground-up in Rust and Apache Arrow. A buffer-agnostic design uses any disk or object store for durability without sacrificing performance. See our benchmarks for concrete numbers vs Debezium and Fivetran.
Do you price by rows or data volume?
Neither. Compute-based pricing scales with cores, not rows. Customers typically see up to 90% cost reduction versus traditional ELT vendors, with 10 to 15 times faster syncs.