Airbyte Self-Hosted
AdvancedData PipelineSelf-hosted Airbyte on your own Kubernetes cluster, with 300-plus connectors and no per-connector cost.
Published 27 September 2026
About Airbyte Self-Hosted
Airbyte Self-Hosted runs the open-source Airbyte platform on your own infrastructure. The lightest documented path is abctl on a single VM, which installs a single-node kind Kubernetes cluster inside the machine; the full path is the official Helm chart on a real cluster, managed or your own. PostgreSQL stores job state and configuration behind the platform.
The connector catalog is the product: more than 300 sources and destinations, from Postgres and MySQL to Salesforce, Stripe, and Snowflake, with a Connector Development Kit for anything missing. Self-hosting runs all of them for a fixed infrastructure bill instead of Airbyte Cloud's volume-based credits.
That saving comes with operations. The platform is Kubernetes-native, so even the single-VM path runs a cluster inside it, which is why this stack carries Kubernetes in core. Upgrades mean re-running the installer or bumping the Helm chart, and failed syncs are debugged through pod logs.
Exposure differs by path. On a cluster it's the ingress controller; an abctl install serves plain HTTP on port 8000, so a TLS-terminating proxy or the cloud's load balancer goes in front before remote access. Teams that only need a handful of scheduled pipelines may be better served by Airflow or Prefect with connection code in the DAG itself.
Key Features
- ✓abctl single-command install on one VM with a bundled kind cluster
- ✓Official Helm chart for full clusters: EKS, GKE, DOKS, or your own
- ✓The full 300-plus connector catalog with no per-connector fee
- ✓PostgreSQL for job state and configuration
- ✓Custom connectors in Python or the Connector Development Kit
When to Use Airbyte Self-Hosted
- →Data teams syncing SaaS sources into a warehouse without volume-based billing
- →Sources on a private network that a hosted service can't reach
- →Organizations whose data must not pass through a third-party pipeline
- →Teams building custom connectors with the Connector Development Kit
- →Platform teams already running Kubernetes that want ELT beside their other workloads
Pros
- Connector breadth is the product, and it ships free in the open-source edition
- Volume-based pricing disappears behind a fixed infrastructure cost
- Runs on any Kubernetes: a single VM through abctl, or a managed or self-managed cluster
Cons
- Kubernetes-native: the lightest path still runs a cluster inside the VM and needs 8 GB of memory
- Upgrades are a recurring maintenance task
- Debugging failed syncs asks for logs and Helm comfort, not just the UI
Hosting Options for Airbyte Self-Hosted
Deploy Airbyte Self-Hosted on Hetzner
The budget path for a single-node abctl install: an 8 GB instance with 4 vCPUs, Airbyte's suggested size, costs around €8-10/mo on a flat-rate root VPS. It carries a handful of concurrent syncs; put a TLS-terminating proxy in front of port 8000 before exposing the UI.
Deploy Airbyte Self-Hosted on DigitalOcean
Two routes: a single 8 GB Droplet for abctl at about $48/mo, or DOKS, a documented Airbyte target with a free control plane, where a pool of 8 GB nodes runs the official Helm chart and a DigitalOcean load balancer handles TLS. The cheapest managed-cluster path in the list.
Deploy Airbyte Self-Hosted on Amazon Web Services
EKS costs about $73/mo for the control plane before nodes, and the Helm chart can move state to RDS for PostgreSQL and logs to S3. It fits when sources and destinations such as RDS or Redshift sit inside the same VPC, so syncs never cross the public internet.
Deploy Airbyte Self-Hosted on Google Cloud Platform
The same Helm chart on GKE, where one zonal cluster pays no management fee, so a small deployment costs only its nodes. BigQuery is a first-class destination, and private networking keeps syncs from Cloud SQL sources inside the project.
These are highlighted picks. To see all the tools, check the Hosting & Cloud category.
Frequently Asked Questions about Airbyte Self-Hosted
Why self-host Airbyte instead of using Airbyte Cloud?
Volume pricing and network reach. Airbyte Cloud's Standard plan starts at $20/mo and bills by data volume through credits, Plus starts at $189/mo, and larger plans price by dedicated capacity. Self-hosting Airbyte Core runs the same connector catalog for a fixed server bill, whatever the volume, and reaches sources on a private network that a hosted service can't. What you give up is managed operations: the cluster, upgrades, and sync failures become yours, and features such as SSO sit in the paid editions.
Why Kubernetes? Can't Airbyte run on one VM?
It can. abctl installs a single-node kind cluster inside one VM and is Airbyte's recommended small deployment; the Helm chart is there for multi-node and managed clusters. Either way the platform is Kubernetes-native, and the VM path hides the cluster inside the machine, which is why it needs more memory than a single container would.
What does the server need, and how many syncs can it run?
Airbyte suggests 4 or more CPUs and 8 GB of memory; low-resource mode runs on 2 CPUs and 8 GB but disables the Connector Builder. Concurrency, not connection count, is the limit: each sync runs as a pod, so one 8 GB node handles a handful of simultaneous jobs and queues the rest. When syncs wait behind each other, the Helm chart on a multi-node cluster is the growth path rather than an ever-bigger VM.
Where does state live, and how do upgrades work?
PostgreSQL holds connections, schedules, and job history. On an abctl install it runs inside the bundled cluster, with data persisted under ~/.airbyte/abctl/ on the host, so back up that directory or dump the database. The Helm chart can point at an external PostgreSQL, such as a managed one, instead. Upgrading abctl means running abctl local install again, optionally pinned with --chart-version; on a cluster it's a Helm chart upgrade. Back up first.
Can I move to Airbyte Cloud later without rebuilding everything?
Not in one click, but not a rewrite either. Connections, destinations, and schedules export through the API or the Terraform provider and re-import on the Cloud side; connector versions and some settings differ between editions. Budget a configuration migration, and expect the sync definitions themselves to move cleanly.
Stacks Related to Airbyte Self-Hosted
Kubernetes DevOps Pipeline
Infrastructure
A production DevOps pipeline on Kubernetes, with a swappable CI/CD engine (GitHub Actions or GitLab CI/CD) and built-in observability.
Strapi Self-Hosted
Infrastructure
Self-hosted Strapi: an open-source headless CMS with PostgreSQL, on a server you control.
Airflow Self-Hosted
Infrastructure
Self-hosted Apache Airflow: the standard data-pipeline scheduler on your infrastructure.
n8n Self-Hosted
Infrastructure
Self-hosted n8n on your own server, with full control over the database, the host, and how it's exposed to the internet.
Scores
Popularity3/5
Airbyte is the best-known open-source ELT tool among data engineers, though less familiar to developers outside data work.
Learning Curve4/5
The UI makes building connections easy, but running the platform means Kubernetes, even on one VM, plus Helm upgrades and pod-level debugging.
Flexibility4/5
More than 300 connectors plus a development kit for custom ones cover almost any source and destination pair.
Performance3/5
Each sync runs as its own pod, so throughput follows the cluster's size; one node handles a handful of concurrent jobs.
Portability3/5
Runs on any Kubernetes and connections export through the API or Terraform, but the Elastic License limits reselling it as a service.
Tools in the Airbyte Self-Hosted Stack
Databases
DevOps & CI/CD
Data Engineering & ETL
Airbyte Self-Hosted Pricing
Airbyte's open-source platform is free under the Elastic License v2: usable freely, but not resellable as a hosted service. The bill is the machine that runs its cluster, an 8 GB single-node abctl install for small sync volumes or a managed cluster at larger scale. Against that sits Airbyte Cloud, from $20/mo on volume-based credits and $189/mo on Plus.
Airbyte suggests 8 GB of RAM with 4 CPUs (2 in low-resource mode) for the bundled kind cluster; sync volumes set the real load.
Managed Kubernetes for the Helm-chart path at larger sync volumes, plus any control-plane fee.
The platform is free under the Elastic License v2: usable freely, not resellable as a hosted service.
Airbyte Self-Hosted System Requirements
source- CPU
- 4 CPUs for best performance; 2 CPUs in low-resource mode
- RAM
- 8 GB
- Platform
- A Kubernetes cluster via abctl (kind inside Docker)
- OS
- Linux (Docker Engine on headless VMs)
The official quickstart's suggested resources; low-resource mode lowers the CPU floor but disables the Connector Builder.