Platform architecture
How the dFlow control plane, worker nodes, networking, builds, and monitoring fit together.
This page uses product terminology: Organisation, Application, Environment, Service, Worker Node, and Deployment. For the conceptual stack, read How dFlow is structured first.
Control plane and worker nodes
- Control plane (dFlow Cloud): Hosted at app.dflow.sh. Where you manage Organisations, Applications, Environments, Services, and releases.
- Worker Nodes: Machines that execute Services for Environments you attach them to (including BYOC). dFlow prepares nodes over SSH (for example Dokku for container-style app deployments). See Worker Nodes concept.
Server access and provisioning
For nodes you connect directly, dFlow uses SSH so you are not required to install a permanent agent on every server. Provisioning depends on your path (dFlow-managed Worker Nodes vs servers you add). Troubleshoot connectivity under Worker Nodes and Compute.
Proxy, DNS, and TLS
Traffic for your Services is routed through the platform proxy layer (Traefik). You assign domains and TLS per Environment. See Domains and SSL under Deployments and Operations.
Builds and deployments
Application Services are typically built as Docker images. dFlow supports Dockerfiles and build tooling such as Railpack where enabled. Builds run in context of your Worker Nodes and Deployments; see Deployments concept.
Scaling and additional compute
To run more capacity, attach additional Worker Nodes to the right Environments or provision managed nodes (provider and regions depend on your plan). You stay in the same Application / Environment / Service model.
Monitoring and logs
Worker Nodes can surface host and workload metrics. The product integrates Netdata for deeper node metrics where installed, alongside built-in views for status and logs. See Logs under Deployments and Operations and Worker Nodes overview under Worker Nodes and Compute.
Advanced: how the dashboard talks to the backend
The control plane UI uses Next.js Server Actions and tRPC over HTTP (plus Payload-backed APIs where applicable). For AI agent access, see dFlow for Agents and dFlow MCP. Broader reference lives under API, CLI, and Admin Reference.