Overview

Choose how to run Refold — Refold Cloud, Refold Managed, or self-hosted — install it, and operate it day to day.

Refold runs as a set of stateless services behind a single load balancer, backed by a database, a cache, and object storage. You decide where those services run and who operates them: fully hosted by Refold, managed by Refold inside your own cloud, or self-hosted end-to-end on your own Kubernetes cluster. This section helps you pick a deployment model, install the platform, and keep it healthy in production.

If you’re new here, start with the deployment models below, then follow the install guide for the option you choose.

Choose a deployment model

Every install is one of three models, which differ in who operates the control plane — the management layer that serves the dashboard and API, stores configuration, and dispatches work. Pick the one that matches your data-residency and operations requirements.

Model	Who operates it	When to pick it
Refold Cloud	Refold, in Refold’s infrastructure	Fastest path — no deployment work.
Refold Managed (BYOC)	Refold, inside your own cloud account	Data residency in your cloud without operating the platform yourself.
Self-Hosted	You, in your own Kubernetes cluster	Full control over upgrades, networking, and the runtime. Highest ops burden.

See the architecture overview for how Refold’s services are structured across each model.

Install

Choose the path that matches your model and environment. Each one brings up the full platform — application services, workers, and their backing stores.

Docker Compose

Run the entire platform on a single host. The quickest way to evaluate Refold.

Helm

Self-host and operate Refold end-to-end on your own Kubernetes cluster.

Refold Managed (BYOC)

Run Refold inside your own AWS account while Refold provisions and manages it.

What you provision

For a self-hosted install you bring the infrastructure Refold runs against. The deployment package can bundle these in-cluster for evaluation, but production should use managed equivalents.

Dependency	Purpose
Database (MongoDB)	System of record — organizations, linked accounts, connectors, workflows, and run history.
Redis	Caching, distributed locks, and the job queues that drive workflow execution.
PostgreSQL	Backs durable workflow orchestration.
Object storage (S3 / GCS / MinIO)	Files, exports, and generated artifacts.

You also need outbound network access to the third-party SaaS APIs your integrations call. See Architecture for the full picture and Benchmarking for sizing.

Configure

Tune a deployment for your environment with the how-to guides:

Set up HTTPS

Serve the dashboard and API over secure connections.

Set up OpenTelemetry

Export traces and metrics for observability.

Configure MinIO

Use MinIO as your object-storage backend.

Operate

Once Refold is running, these guides cover day-2 operations:

Backup & restore

Protect your data and recover from a snapshot.

Troubleshooting

Diagnose and fix common deployment issues.