Airbyte

Name: Airbyte
Brand: Airbyte

Open-source data integration with 600 plus connectors and a build-your-own kit

Airbyte is an open-source data integration platform that moves data from hundreds of sources, SaaS applications, databases, APIs, and files, into warehouses, lakes, and other destinations. It offers a self-hosted open-source edition, a managed cloud service, and a connector development kit that lets teams build a connector for any system in hours rather than waiting for a vendor to support it.

Visit website

Overview

The ELT market before Airbyte had a structural problem: proprietary vendors supported the connectors that were commercially worthwhile, and everything else was a custom script somebody maintained in their spare time. Airbyte's answer was to open-source the connector layer and make building new ones cheap, which produced the largest connector catalogue in the category and, more importantly, a path for the long tail nobody else supports.

For go-to-market teams the relevance is direct. The warehouse-native architecture that reverse ETL and composable CDP tools depend on requires data to arrive in the warehouse first: CRM records, ad platform spend, support tickets, billing history, product events. Airbyte is one of the standard ways that happens, and its pricing model, based on volume synced rather than per-connector, makes broad coverage affordable for smaller teams.

The honest caveats concern operational maturity. Community-maintained connectors vary in quality and a long-tail connector can break when a source API changes, self-hosting requires real infrastructure work, and cloud pricing based on data volume can surprise teams syncing large tables frequently. Certified connectors and the managed service address most of this, and the trajectory has been steadily toward reliability, but a team choosing Airbyte should still evaluate the specific connectors it depends on rather than the catalogue size.

Best for

Data teams that need broad source coverage including long-tail systems, want the option to self-host or build connectors, and prefer volume-based pricing over per-connector licensing.

Not the right fit for

Teams with no data engineering capacity, where a fully managed proprietary service with a narrower but hardened catalogue is safer.
Organizations that need guaranteed connector reliability across the long tail, since community connectors vary.
Businesses without a warehouse, since Airbyte moves data into one rather than analyzing it.
Reverse ETL use cases; Airbyte primarily loads into warehouses rather than activating data outward.
Very small operations where a handful of native integrations already cover every need.

How it works

1
You choose a source and a destination from the connector catalogue, authenticate both, and select which tables or streams to replicate along with a sync mode: full refresh, incremental append, or incremental deduplicated history.
2
Airbyte extracts from the source and loads raw or normalized data into the destination on a schedule, tracking state so incremental syncs resume rather than repeating work. Failures are retried, and sync history is visible per connection.
3
Transformation is deliberately left downstream, following ELT convention: land the data first, then model it with dbt or SQL in the warehouse. Airbyte can trigger dbt jobs after a sync so modelling runs against fresh data.
4
When a needed source has no connector, the connector development kit and low-code builder let a team create one, either keeping it private or contributing it back. This is the escape hatch that distinguishes the platform from closed alternatives.

Feature breakdown

20 features in 4 modules

Connectors

The catalogue and the ability to extend it.

Six hundred plus connectors: The broadest catalogue in the category, spanning SaaS applications, databases, APIs, files, and cloud storage.
Connector development kit: Build a connector in Python or through a low-code builder when a source is unsupported, rather than waiting for a vendor roadmap.
Certified versus community tiers: Vendor-maintained certified connectors carry reliability commitments, while community connectors cover the long tail with more variability.
Database replication: Change data capture from Postgres, MySQL, SQL Server, and others for near-real-time database replication.
Custom API sources: The low-code builder turns a documented REST API into a working connector without a full development project.

Sync engine

Moving data reliably at volume.

Incremental sync modes: Append, deduplicated history, and full refresh options per stream, with state tracking so runs resume correctly.
Schema change handling: Detection and configurable propagation of source schema changes, so a new column does not silently disappear.
Scheduling and orchestration: Cron-style scheduling plus API and orchestrator triggering from Airflow, Dagster, or Prefect.
Retry and failure handling: Automatic retries with detailed logs per sync attempt for diagnosing source-side failures.
dbt integration: Trigger transformation jobs after loads so modelling always runs against fresh data.

Deployment options

Self-host, cloud, or in your own infrastructure.

Open-source self-hosting: Run the full platform yourself on Kubernetes or Docker with no license fee and complete data control.
Airbyte Cloud: Fully managed service with usage-based pricing and no infrastructure to operate.
Self-managed enterprise: Commercially supported deployment inside your own infrastructure with enterprise features and support.
Terraform provider and API: Connections managed as code, which keeps pipeline configuration under normal engineering review.
Multi-workspace management: Separate environments and teams under one deployment, useful for agencies and larger organizations.

Governance and operations

Running it as production infrastructure.

Sync monitoring and alerting: Failure notifications and run history so broken pipelines are noticed before a dashboard is wrong.
Column selection and hashing: Choose which fields to replicate and hash sensitive columns in transit, limiting what lands in the warehouse.
Role-based access control: Permissions over workspaces and connections on enterprise deployments.
Audit logging: Records of configuration changes for compliance and troubleshooting.
Compliance certifications: SOC 2 and GDPR posture on the cloud service, with self-hosting available where data cannot leave your environment.

Use cases

4 documented

Data engineer consolidating go-to-market data

CRM, ad platforms, billing, and support data live in separate systems and reporting is assembled by hand each month.

Connectors land all of it in the warehouse on a schedule, dbt models unify it, and the monthly assembly ritual disappears.

Analytics team blocked by an unsupported source

A niche vertical application holds critical data and no ELT vendor supports it.

A connector is built with the low-code builder in a day, and the data joins the same pipeline as everything else.

Startup avoiding per-connector pricing

Proprietary alternatives quote by connector and by rows, making broad coverage expensive at low volumes.

Self-hosted Airbyte covers a dozen sources with no license cost, at the price of running the infrastructure.

Regulated organization with residency requirements

Customer data cannot pass through a third-party processor outside its own environment.

Self-managed deployment inside the company's own infrastructure keeps data in place while still providing managed-style tooling.

Pricing

from Free self-hosted; Cloud usage-based with a trial credit, commonly from tens of dollars per month at small volumes

Open-source edition is free to self-host. Airbyte Cloud is usage-based, priced by data volume synced with different rates for API sources and database replication. Self-managed enterprise is a quoted annual subscription.

Plan	Price	Includes
Open source	$0 self-hosted	Full connector catalogue and sync engine Complete data control and no license cost Infrastructure and maintenance are your responsibility
Cloud	Usage-based monthly, by volume synced	Fully managed with no infrastructure to run Certified connectors with reliability commitments Monitoring, alerting, and standard support
Self-managed enterprise	Quoted annual	Commercial support for deployment in your own infrastructure Role-based access control, audit logging, and SSO Enterprise SLAs and dedicated support

Billing notes

Cloud pricing is volume-based, so large tables synced frequently cost considerably more than small ones synced daily; review sync frequency deliberately.
Database replication and API source volumes are priced differently, which matters when a single large database dominates usage.
Self-hosting has no license cost but real infrastructure and maintenance cost; budget engineering time, not zero.
Warehouse storage and compute for landed data is separate and paid to your cloud provider.
Prices and rate structures as published August 2026; usage-based pricing in this category has changed repeatedly across vendors.

Value assessment: Airbyte's economics are strongest where proprietary ELT is weakest: many sources, some of them obscure, at moderate volume. Self-hosting removes license cost entirely for teams with infrastructure capability, and the connector kit removes the roadmap dependency that makes closed vendors frustrating. The counter-argument is reliability: a certified connector on a managed service is a different proposition from a community connector you now maintain. Judge value by the specific sources you need rather than by catalogue size.

Strengths & limitations

Strengths

The largest connector catalogue in the category by a wide margin.
Connector development kit removes the vendor roadmap dependency for unsupported sources.
Genuine open-source self-hosting for teams with residency requirements or cost constraints.
Volume-based pricing rather than per-connector licensing suits broad, shallow coverage.
Configuration as code through Terraform and API keeps pipelines under engineering review.
Strong integration with the surrounding modern data stack including dbt and orchestrators.

Limitations

Community connector quality varies, and long-tail connectors can break when source APIs change.
Self-hosting is a real operational commitment, not a free lunch.
Usage-based cloud pricing can escalate with large or frequently synced tables.
Loads into warehouses rather than activating data outward, so a separate reverse ETL tool is needed.
Less hardened than long-established proprietary alternatives on the most critical enterprise sources.
No transformation layer by design, so dbt or equivalent is a prerequisite for usable data.

Head-to-head comparisons

3 alternatives

Airbyte vs Fivetran

from Free for up to around 500,000 monthly active rows; paid usage from roughly $500 per month at modest volumes

The defining comparison in ELT. Fivetran offers fewer connectors but hardens them thoroughly, with automatic schema handling and a reliability reputation that justifies its premium for critical pipelines. Airbyte offers far broader coverage, open source self-hosting, and a connector kit for anything missing. Teams with budget and a short list of critical sources often prefer Fivetran; teams with many sources, tighter budgets, or engineering capacity prefer Airbyte.

Full Airbyte vs Fivetran comparison

Airbyte vs Hightouch

from Free tier for limited syncs; paid plans commonly from several hundred dollars per month

Opposite directions of the same architecture. Airbyte loads data into the warehouse, Hightouch activates modelled data out to business tools. They are complements rather than alternatives, and the pair plus dbt is close to a standard stack for warehouse-native go-to-market teams.

Full Airbyte vs Hightouch comparison

Airbyte vs Dataddo

from Free for a small number of flows; paid plans from roughly $99 per month

Dataddo targets less technical buyers with a fully managed, no-code service that can also write to dashboards and business tools directly, at predictable per-flow pricing. Airbyte targets engineers, with broader coverage, self-hosting, and extensibility. Small marketing teams without a data function are usually better served by Dataddo; data teams almost always prefer Airbyte.

Full Airbyte vs Dataddo comparison

Implementation & onboarding

Setup time: Cloud connections take minutes each once credentials exist. Self-hosted deployment on Kubernetes takes days to do properly, and a complete pipeline including modelling takes weeks.
Learning curve: Moderate for data engineers, steeper for anyone unfamiliar with ELT concepts such as sync modes and incremental state. Building custom connectors requires real development skill even with the low-code builder.
Onboarding: Self-serve with extensive documentation and a large community. Enterprise agreements include implementation support and connector reliability commitments.
Migration notes: Moving from another ELT tool means recreating connections and, importantly, reconciling destination schemas, since vendors normalize differently. Plan a parallel run and validate row counts before decommissioning the previous pipeline; historical backfills should be scheduled deliberately because they dominate volume-based cost.

Platform, API & security

Platforms: Docker and Kubernetes self-hostingAirbyte CloudTerraform providerREST API
API: Full REST API and Terraform provider for managing sources, destinations, and connections, plus a connector development kit in Python and a low-code builder.
Compliance: GDPRCCPASOC 2 Type IIHIPAA support on enterprise plans
Data residency: Complete control when self-hosted; regional options on the cloud service.
SSO: Available on enterprise deployments.
Security notes: Column selection and hashing limit what sensitive data reaches the destination, and self-hosting keeps data entirely within your environment, which is often the deciding factor for regulated buyers.

Support & resources

Channels: Community forum and SlackTicket support on CloudDedicated support on enterprise plans
Documentation: Comprehensive documentation covering deployment, connectors, sync modes, and the development kit, with per-connector reference pages.
Community: One of the largest open-source data engineering communities, which is also the mechanism by which the long-tail connector catalogue is maintained.

Company

Founded: 2020
Headquarters: San Francisco, California, United States
Ownership: Private, venture-backed
Founders: Michel Tricot, John Lafleur
Employees: ~200 (est. 2026)
Funding: Raised substantial venture funding including a Series B round at a high valuation.

Timeline

2020Founded with an open-source approach to ELT connectors, targeting the long tail proprietary vendors ignore.
2021Raises rapid follow-on funding as the open-source connector catalogue expands quickly.
2022Launches Airbyte Cloud, adding a managed option alongside self-hosting.
2023Introduces the low-code connector builder, lowering the effort to support a new API.
2025Focuses on connector reliability tiers and enterprise self-managed deployments.
2026Established as the open-source standard for data integration with the broadest connector coverage available.

Integrations

Snowflake
BigQuery
Databricks
Redshift
Postgres
Salesforce
HubSpot
Stripe
Google Ads
Meta Ads
dbt
Airflow

Frequently asked questions

10 questions

What is Airbyte?

Airbyte is an open-source data integration platform that moves data from hundreds of sources into warehouses, lakes, and other destinations. It is available as free self-hosted software, a managed cloud service, and a self-managed enterprise edition, and includes a development kit for building connectors to unsupported systems.

Is Airbyte really free?

The open-source edition has no license fee and can be self-hosted indefinitely. It is not free of cost: you pay for the infrastructure it runs on and the engineering time to operate it. Airbyte Cloud is the paid managed alternative, priced by volume synced.

Airbyte vs Fivetran: which should I choose?

Fivetran has fewer connectors but hardens them heavily and has a strong reliability reputation, at a premium price. Airbyte has far broader coverage, open-source self-hosting, and a connector kit for anything missing. Choose Fivetran for a small number of business-critical pipelines with budget available; choose Airbyte for breadth, cost control, or extensibility.

How much does Airbyte Cloud cost?

It is usage-based, priced on the volume of data synced, with different rates for API sources and database replication. Small deployments can run at tens of dollars a month, while large frequently synced databases cost substantially more. Review sync frequency and table selection carefully, since both drive the meter directly.

Can Airbyte connect to a system that has no connector?

Yes, through the connector development kit or the low-code builder, which turns a documented REST API into a working connector without a full development project. This is the main structural advantage over closed platforms, where an unsupported source means waiting on a vendor roadmap.

Does Airbyte transform data?

Only minimally. It follows the ELT pattern: land raw data in the warehouse, then transform it there with dbt or SQL. Airbyte can trigger dbt jobs after a sync so modelling runs on fresh data, but the transformation logic itself lives in your own models rather than in the pipeline.

Are community connectors reliable?

It varies, which is the honest answer. Certified connectors carry vendor maintenance and reliability commitments, while community connectors cover the long tail with more variability and can break when a source API changes. Check the tier of the specific connectors you depend on rather than judging by catalogue size.

Can Airbyte do reverse ETL?

It is primarily built to load data into warehouses, and while some destinations write to business tools, activation is not its focus. Teams wanting to sync modelled warehouse data into CRM, ads, and messaging tools should pair it with a reverse ETL platform such as Hightouch or Census.

Should I self-host or use the cloud service?

Self-host if you have Kubernetes capability and either strict data residency requirements or enough volume that usage pricing hurts. Use the cloud service if your team's time is better spent on modelling than on operating pipelines, which is true for most organizations below serious data engineering scale.

Does Airbyte support change data capture from databases?

Yes, for common databases including Postgres, MySQL, and SQL Server, replicating changes rather than repeatedly reading whole tables. This is both the efficient way to keep a warehouse current and, on the cloud service, the usage pattern that most affects cost, so configure it deliberately.

Editorial verdict

Airbyte fixed the structural flaw in commercial ELT: that vendors only support connectors worth their while, leaving everything else to homemade scripts. Open-sourcing the connector layer produced both the broadest catalogue in the category and, more valuably, an escape hatch for the source nobody else supports. Self-hosting gives residency-constrained and cost-sensitive teams a genuine option, and volume-based cloud pricing beats per-connector licensing for broad shallow coverage. The trade is variance: long-tail connectors are maintained by the commons, self-hosting is real work, and usage pricing punishes large tables synced too often. Evaluate the connectors you actually need, not the number on the homepage, and it is a strong default for any team with a data function.

Written by the SaaSTracker editorial team. Awards, when shown, are judged against the published criteria in our methodology.