Skip to main content
Open source · PostgreSQL · Real-time · Scale-out · Iceberg Firebolt is built for engineers and AI agents: low-latency queries, efficient ELT, and one engine that behaves the same from a laptop to a production cluster.

​
Why Firebolt

  • Open engine, open storage. The engine is open source and reads open formats (Parquet, Iceberg), so there is no lock-in.
  • Deploy anywhere. The same binary runs on a laptop, VM, bare metal, Kubernetes, or any cloud. Managed and bring-your-own-cloud (BYOC) commercial offerings are also available.
  • Price-performance. Fast analytics on data-intensive workloads at low cost.
  • Real-time and batch. Low-latency queries on the custom storage format and batch ELT over Parquet and Iceberg, from one engine.

​
Deployment modes

Single binary. Zero dependencies for local development and embedded use. Install with: 
bash <(curl -s https://get-core.firebolt.io/)
Helm chart. Scale compute and storage independently to isolate workloads on Kubernetes. See the Helm chart quickstart. Kubernetes operator. Run a full platform with workload isolation and dynamic scaling. See the Firebolt Operator quickstart. Managed service. Start on the fully managed service with a $200 credit and no card. See Managed service. For the architecture behind each mode, see Architecture.

​
At a glance

AttributeDetail
SQL dialectPostgreSQL-compatible
Data volumeGigabytes to petabytes
ArchitectureSingle binary, built for object storage
Computex86 / ARM, 1 to 256 vCPU, 1 to 256 nodes
ConnectivityADBC, JSON REST, PostgreSQL wire protocol
StorageCustom format (real time), Parquet and Iceberg (batch)
ConsistencyACID transactions, snapshot isolation
DeploymentBinary, Docker, Helm, Kubernetes operator, managed service
CloudAWS (GA), GCP (GA), Azure (Preview)

​
Next steps