Estuary Flow
Estuary Flow is a data movement platform built on a streaming core, capturing from databases via CDC and from SaaS APIs, then materialising continuously into warehouses and lakes rather than on a schedule.
LicenseSource-available· BUSL-1.1
DeploymentManagedSelf-hostedHybrid
PricingFreeSubscription
Operational complexityLow
WorkloadBatchStreaming
LanguageRust
Use it when
- The same pipeline must feed a warehouse on a schedule and a consumer that needs changes within seconds.
- Database CDC is the main source and low-latency, log-based capture matters.
- New destinations should backfill from retained collection history instead of re-reading the source.
- You want 200+ managed connectors without operating streaming infrastructure yourself.
Think twice when
- Everything is genuinely batch; a schedule-driven tool like Airbyte or Fivetran is simpler to reason about.
- Per-GB pricing plus a monthly per-connector fee meets sustained high volume.
- Open-source licensing matters: the engine is mostly Business Source License, source-available rather than open source.
- Transformations beyond SQL, TypeScript, or Python derivations belong in a downstream framework.
How it runs
Estuary operates the data plane as a managed service, with private deployment and BYOC as options. Captures run continuously, collections persist as JSON in your own cloud storage bucket, and materializations keep destinations current within milliseconds or on a per-destination schedule. There is no scheduler to run: pipelines are continuous, billed at $0.50 per GB moved plus a monthly connector fee.
Details
How Estuary Flow answers the questions each capability turns on.
Batch ELT
| How it works | |
| Authoring model | Low-code UI, or flow.yaml specs via the flowctl CLI |
| Incremental sync | Incremental by design; collections keep history for re-backfills |
| Sync frequency | Continuous, or a sync schedule per materialization, 30 min default |
| Schema drift | Schema inference; autoDiscover adds fields, breaking changes backfill |
| Running it | |
| Scheduling | None needed: pipelines run continuously, not on a scheduler |
| Connections | |
| Source coverage | Databases, SaaS APIs and object storage, all as captures |
| Destinations | Warehouses, lakes, streams and SaaS targets as materializations |
| Custom connectors | Open Docker-image protocol; any language |
| Transformations | Derivations in SQL, TypeScript or Python; dbt Cloud after load |
| Cost | |
| Billing unit | $0.50 per GB moved plus a monthly fee per connector |
Change Data Capture
| How it works | |
| Initial snapshot | Backfill from collection history, not from the source |
| End-to-end lag | Sub-100ms end to end; streaming rather than polled |
| Delivery guarantee | Exactly-once into transactional targets · else At-least-once |
| Schema drift | autoDiscover evolves the collection; breaking changes re-version it |
| Running it | |
| What you run | Nothing: Estuary runs the data plane; private or BYOC optional |
| Connections | |
| Sources | 200+ connectors: databases and SaaS sources |
| Source requirements | Replication slot and publication on Postgres; ROW binlog on MySQL |
| Transformations | Derivations: continuous SQL, TypeScript or Python over collections |
| Destinations | 200+ materializations; collections in your own bucket |
| Cost | |
| Billing unit | $0.50 per GB moved plus a monthly fee per connector |
Alternatives to Estuary Flow
