Integration & ETL Plane · ELT & Ingestion

Airbyte

Data integration platform with a large connector catalogue, self-hosted or as a cloud service.

Overview

Airbyte is the open-core answer in this category, and the reason most teams look at it is simple: you can run the whole platform yourself, free, on Kubernetes, with no volume cap.

Be precise about the licence though, because "open source" overstates it. The repository is dual-licensed, with connectors and the development kit open source and the core platform under the Elastic License 2.0, which is source- available and forbids offering the product as a managed service.

Its catalogue is the largest here, though the count varies by page between 600 and 700, with the live catalogue rendering 653. Cite the page rather than the number.

What makes that catalogue usable is a published four-tier support model. Airbyte-maintained connectors carry support and upgrade windows; marketplace connectors are community-maintained with no service level and an explicit warning that they may break without notice.

Building your own is unusually accessible. The Connector Builder is a no-code interface needing no local development environment, backed by a low-code YAML kit and a full Python kit for anything harder.

Transformation is deliberately thin. Airbyte types and deduplicates on load and integrates dbt for cloud users; it is a movement tool, not a modelling one.

The 2025 architectural move was Enterprise Flex, splitting a managed control plane from data planes you run yourself, so records and credentials never leave your environment. That is also how it answers HIPAA and data residency, by architecture rather than by certification.

The 2026 repositioning is towards AI, describing itself as a context layer for agents, with an agents product and context store, and vector databases as first-class destinations.

One gap worth knowing when comparing: there is no published equivalent of column blocking or hashing, and no type 2 history mode.

Key features and capabilities

The same headings are used for every elt & ingestion entry, so two tools can be read side by side.

Connectors and coverage
  • Counts differ by page, 700 on the connectors page against 653 rendered in the live catalogue
  • A published four-tier support model, Airbyte-maintained, enterprise, marketplace and custom, plus archived
  • Marketplace connectors are community-maintained with no service level and may break without notice
  • Enterprise connectors cost extra, are Enterprise or Pro only, and add parallelism and service levels
How pipelines are built
  • A connection pairs a configured source with a configured destination, built in the interface or as code
  • Code paths are the REST API, an official Terraform provider and PyAirbyte as a Python library
  • Scheduling by interval, cron expression or manual trigger; no minimum frequency is published
  • Transformation is limited to typing and deduplication, with dbt integration for cloud users
Loading and sync behaviour
  • Five published sync modes combining full refresh or incremental reads with overwrite, append or dedupe writes
  • Cursor-based incremental is supported by any connector whose source allows it
  • Change data capture is an alternative method, supported only by some sources, which are not enumerated
  • No type 2 history or slowly changing dimension support is published
Destinations
  • Warehouses including Snowflake, BigQuery, Redshift and Databricks
  • Object storage on S3, Google Cloud Storage and Azure Blob Storage
  • Both Iceberg and Delta are supported as table formats
  • Vector databases are first-class destinations, a deliberate retrieval and AI play
Security, monitoring and lineage
  • SOC 2 Type II and ISO 27001, the latter certified in November 2022
  • HIPAA and GDPR are answered architecturally through customer-hosted data planes, not as certifications
  • Encryption in transit throughout, with cloud metadata encrypted at rest and secrets in Google Secret Manager
  • Role-based access control in the cloud only; the open-source edition has none and relies on network controls
  • Audit logs on Pro and Enterprise Flex, retained 365 days; column blocking and hashing are not published
Extending it yourself
  • The Connector Builder creates API sources with no local development environment, but no destinations
  • A low-code YAML kit for HTTP APIs, allowing custom Python components, and a full Python kit beyond that
  • PyAirbyte packages connectors as a Python library, usable inside Dagster or Airflow
  • An official Terraform provider, available to open-source, cloud and enterprise users
Where it runs and what it costs
  • Self-managed community edition, free and self-hosted on Kubernetes, locally or across a cluster
  • Cloud in Standard, Plus and Pro tiers, with control planes in the United States
  • Enterprise Flex runs a managed control plane with your own data planes, including on-premises and air-gapped
  • Data planes available in the United States and the European Union, one region per workspace

Pricing

Usage-basedFree self-hosted; cloud from $20 a month

Self-hosting the community edition is free with no volume cap, subject to the Elastic Licence. Cloud Standard starts at $20 a month with five credits included and further credits at $5 each. Plus starts at $189 a month for 40 credits, with bundles down to roughly $3.20 a credit at 1,000 and overage at $5. Pro and Enterprise Flex move to capacity pricing measured in data workers, and no rate is published for either, so both are quote-only. The trial runs 30 days with 400 credits and no card.

Vendor pricing page →

Demos and videos

About Airbyte

Airbyte was founded in 2020 by Michel Tricot and Jean Lafleur, going through Y Combinator's winter 2020 batch, and raised $181m across a $5.2m seed led by Accel, a $26m Series A and a $150m Series B in December 2021 that made it a unicorn. It is private, with no announced acquisitions. It publishes adoption rather than revenue: 7,000 companies, a fifth of the Fortune 500, more than 27,000 community members and two million pipelines synced daily. The licence split is the thing to check: connectors and the development kit are open source, while the platform is Elastic License 2.0.

Founded 2020 · San Francisco, California · airbyte.com

Other elt & ingestion tools

AWS Glue

Integration & ETL Plane · ELT & Ingestion

Serverless data integration service on AWS for discovering, preparing and moving data.

  • Cloud service

Azure Data Factory

Integration & ETL Plane · ELT & Ingestion

Azure's managed service for data integration and ETL/ELT pipelines.

  • Cloud service

Data Factory in Microsoft Fabric

Integration & ETL Plane · ELT & Ingestion

Pipelines and Dataflow Gen2 for ingesting and transforming data in Microsoft Fabric.

  • Cloud service

dlt

Integration & ETL Plane · ELT & Ingestion

Open-source Python library for loading data from APIs and files into warehouses and lakes.

  • Open source

Fivetran

Integration & ETL Plane · ELT & Ingestion

Managed ELT connectors that replicate data from SaaS applications and databases into warehouses and lakes.

  • Commercial

Matillion

Integration & ETL Plane · ELT & Ingestion

Cloud data integration and transformation platform for building pipelines on cloud data platforms.

  • Commercial

Drafted with AI assistance and checked against the vendor’s own documentation.