Estuary vs Airbyte

Quick verdict

Estuary was formerly named Estuary Flow; the Flow name survives in its GitHub repository and the flowctl CLI. A capture ingests data into collections, JSON documents stored in a cloud storage bucket and validated against a JSON schema, and a materialization keeps a destination updated as new documents arrive. Databases are captured with log-based CDC; SaaS sources and files with scheduled batch loads. Transformations, called derivations, run in SQL (SQLite) or TypeScript, and Python on private and BYOC data planes.

Airbyte is a data movement platform with a large connector catalogue. Its CDC documentation covers PostgreSQL, MySQL, SQL Server, MongoDB, Oracle, SAP HANA and IBM Db2. Airbyte Core is free to self-host on Kubernetes under the Elastic License 2.0; Airbyte Cloud (Standard, Plus, Pro) and Enterprise Flex are paid. For Airbyte as a general ELT tool, see Airbyte vs Fivetran; this page looks only at database CDC and replication behaviour.

Side by side

Key differences

Continuous capture versus scheduled CDC syncs

This is the main behavioural difference. Estuary holds a connection to the database log and writes changes into collections continuously. Destinations are kept up to date by materializations, and the PostgreSQL connector documentation lists real-time capture while noting that destination sync cadence can still be controlled on the materialization side. Because collections are stored once, the same capture can feed a warehouse, a search index and a Kafka-compatible consumer.

Airbyte runs CDC as a scheduled sync. Each run reads the change log from the saved position up to the time the sync started and then stops; Airbyte's CDC page says CDC sources are not treated as infinite streaming sources, so the schedule must be frequent enough to keep up with log growth. The fastest cloud schedule is every 5 minutes on Pro and Enterprise Flex, 15 minutes on Plus and 1 hour on Standard. In our view, if minutes of delay are acceptable, Airbyte's model is simpler to reason about; if downstream systems need changes within seconds, Estuary's model fits better.

PostgreSQL CDC in practice

Both use PostgreSQL logical replication with a replication slot and a publication, so the database setup looks similar. Airbyte's Postgres connector requires the pgoutput plugin and a slot used only by Airbyte, and now requires PostgreSQL 14 or later for CDC because the connector moved to Debezium 3.4. Estuary's connector supports PostgreSQL 10 and later, needs one slot per capture and uses a small watermarks table (public.flow_watermarks by default) to keep backfills accurate. A typical setup on the source looks like this:

-- PostgreSQL: objects a log-based CDC tool needs (names vary by tool)
CREATE PUBLICATION cdc_publication FOR TABLE public.orders, public.customers;
SELECT pg_create_logical_replication_slot('cdc_slot', 'pgoutput');

-- Watch slot lag so WAL does not pile up between reads
SELECT slot_name, active,
       pg_size_pretty(pg_wal_lsn_diff(pg_current_wal_lsn(), restart_lsn)) AS retained_wal
FROM pg_replication_slots;

The slot retains WAL until the tool reads it. With a continuous reader the slot advances all the time; with scheduled syncs it advances only when a sync runs. Airbyte warns that if max_slot_wal_keep_size is exceeded PostgreSQL can invalidate the slot, after which the connection needs a full re-sync. Both tools support capture from a read replica only on PostgreSQL 16 and later.

Deletes and table state in the destination

Estuary marks delete events with _meta/op set to d. By default destinations keep soft-deleted rows, which you can filter out on that field; hard deletes, which physically remove the row, are supported on destinations including Snowflake, BigQuery, Databricks, Redshift, PostgreSQL, MySQL, SQL Server and MongoDB, but not on bindings that use delta updates. Airbyte delivers inserts, updates and deletes in the same record format, sets _ab_cdc_deleted_at only on deletes, and notes that the final table will not show rows whose latest entry is a delete. Decide early whether analysts need deleted rows kept for history, and configure each tool to match.

Running it and paying for it

Airbyte can be run entirely on your own Kubernetes cluster as Airbyte Core, at no licence cost but with the operations work that implies; the cloud plans and Enterprise Flex remove that work. Estuary has no self-hosted edition: it is a managed service, with private data planes or BYOC (a data plane in your own AWS, GCP or Azure account) on the Enterprise plan. For licensing, Airbyte's platform is ELv2 and Estuary's core repository is under the Business Source License; neither is OSI open source.

Pricing works differently. Estuary charges per GB, counting data both when captured and when delivered, plus a monthly fee per connector instance, so one capture feeding two destinations adds GB and instance fees. Airbyte Cloud charges credits by volume on Standard and Plus, or capacity in Data Workers on Pro and Enterprise Flex, which does not rise with volume. Model both with your real change volume and the number of sources and destinations.

Pricing and licensing

Estuary. Listed on estuary.dev/pricing on 7 November 2026 (USD): Developer is free up to 10 GB a month and 2 connector instances, with no credit card. Cloud is pay-as-you-go at USD 0.50 per GB plus USD 100 a month per connector instance (USD 50 after the sixth), prorated hourly, with a 30-day trial when you exceed Developer limits. Data is counted each time it is captured and each time it is delivered; Estuary's own example is 100 GB a month from Postgres to Snowflake counted as 200 GB (USD 100) plus 2 connector instances (USD 200), USD 300 a month. Enterprise (private and BYOC deployment, SSO, SOC 2 report, HIPAA BAA) is custom-priced on annual contracts.

Airbyte. Listed on airbyte.com/pricing on 7 November 2026 (USD): Standard starts at USD 20 a month with 5 credits, billed by volume; Plus is sold as monthly credit packages from USD 189 (40 credits) to USD 4,999 (2,000 credits); Pro and Enterprise Flex are capacity based (Data Workers) and quote-only. Standard and Plus have a 14-day trial. Airbyte Core is free to self-host; you pay for the infrastructure.

Pricing checked on the vendors' official pages on 7 November 2026. Prices change; confirm before buying.

Where each one leads

Limitations

When to choose each

When neither is right

• You want open source CDC into Kafka that you run yourself: see Debezium vs Oracle GoldenGate.
• You need enterprise sources such as Db2 for z/OS or SAP applications with self-hosted replication: see Fivetran HVR vs Qlik Replicate.
• You need CDC plus streaming SQL processing from a vendor: see Qlik Replicate vs Striim.
• You mainly load SaaS data on a schedule rather than database changes: see Airbyte vs Fivetran.

Final recommendation

Frequently asked questions

Sources

• Estuary pricing
• Estuary concepts
• Estuary PostgreSQL capture connector
• Estuary deletions reference
• Airbyte pricing
• Airbyte change data capture
• Airbyte Postgres source connector
• Airbyte licences

Checked November 2026.

How we research comparisons: our editorial method.