Skip to main content


A capture is how Flow ingests data from an external source. Every Data Flow starts with a capture.

Captures are a type of Flow task. They connect to an external data source, or endpoint, and bind one or more of its resources, such as database tables. Each binding adds documents to a corresponding Flow collection.

Captures run continuously: as soon as new documents are made available at the endpoint resources, Flow validates their schema and adds them to the appropriate collection. Captures can process documents up to 16 MB in size.

You define and configure captures in Flow specifications.

See the guide to create a capture

Pull captures

Pull captures pull data from an endpoint using a connector.

Estuary sources

Estuary builds and maintains many real-time connectors for various technology systems, such as database change data capture (CDC) connectors.

See the source connector reference documentation.

Airbyte sources

Flow also natively supports Airbyte source connectors. These connectors tend to focus on SaaS APIs, and do not offer real-time streaming integrations. Flow runs the connector at regular intervals to capture updated documents.

Airbyte source connectors are independently reviewed and sometime updated for compatibility with Flow. Estuary's source connectors documentation includes actively supported Airbyte connectors. A full list of Airbyte's connectors is available at Airbyte docker hub. If you see a connector you'd like to prioritize for access in the Flow web app, contact us.


To help you configure new pull captures, Flow offers the guided discovery workflow in the Flow web application.

To begin discovery, you tell Flow the connector you'd like to use and basic information about the endpoint. Flow automatically generates a capture configuration for you. It identifies one or more resources — tables, data streams, or the equivalent — and generates bindings so that each will be mapped to a data collection in Flow.

You may then modify the generated configuration as needed before publishing the capture.


Pull captures are defined in Flow specification files per the following format:

# A set of captures to include in the catalog.
# Optional, type: object
# The name of the capture.
# Endpoint defines how to connect to the source of the capture.
# Required, type: object
# This endpoint uses a connector provided as a Docker image.
# Docker image that implements the capture connector.
# File that provides the connector's required configuration.
# Configuration may also be presented inline.
config: path/to/connector-config.yaml

# Bindings define how collections are populated from the data source.
# A capture may bind multiple resources to different collections.
# Required, type: array
- # The target collection to capture into.
# This may be defined in a separate, imported specification file.
# Required, type: string
target: acmeCo/example/collection

# The resource is additional configuration required by the endpoint
# connector to identify and capture a specific endpoint resource.
# The structure and meaning of this configuration is defined by
# the specific connector.
# Required, type: object
stream: a-bucket/and-prefix
# syncMode should be set to incremental for all Estuary connectors
syncMode: incremental

- target: acmeCo/example/another-collection
stream: a-bucket/another-prefix
syncMode: incremental

Push captures

Push captures expose an endpoint to which documents may be pushed using a supported ingestion protocol.


Push captures are under development. Estuary intends to offer Webhook, Websocket, and Kafka-compatible APIs for capturing into collections. Join the Estuary Slack for more information on this and other ongoing development work.


Push capture configurations use the following general format:


# The name of the capture.
# This endpoint is an ingestion.
ingest: {}

- # The target collection to capture into.
target: acmeCo/example/webhooks
# The resource configures the specific behavior of the ingestion endpoint.
name: webhooks