Services, workloads, and placement

Connect an existing service so Faros can reach it through an edge, or deploy a container workload to connected Kubernetes clusters.

Prerequisites

Sign in to Faros SaaS and select a workspace with Edges enabled. You need permission to create services or workloads and a connected edge . Services can use a Linux server or Kubernetes edge; workloads require a Kubernetes edge. Have the application’s address and any required credential ready. For workloads, confirm the image can be pulled by the target cluster and inspect its edge labels before selecting targets.

Connect a service

1. Open Edges → Services and choose Create service.
2. Enter a Name, select the Edge, and choose the service Type.
3. Set Scheme and Port. A preset may fix the scheme.
4. For Host / IP, enter the address reachable from that edge, not from your laptop. For Kubernetes Service, enter Target namespace and Target service name. Linux server edges use the host option.
5. Choose Create service, then open the service and configure its credentials if the selected type requires them.
6. Check its validation result and reachability before using it with an assistant.

Expect the service to be listed on the selected edge and validation to succeed. Then use MCP setup to connect an assistant and inspect the tools exposed for that service. A listing alone does not verify connectivity or access.

Available service presets

Edges supports the following service types. Presets describe how to connect to an existing application; they do not install it. The application must be reachable from the selected edge, and its credential must permit the operations you intend to expose.

Inspect the service schema for connection fields and the service catalog for preset operations and authentication. Home Assistant has a dedicated tool implementation . A generic service does not automatically gain every preset’s tools. After validation, use the MCP client’s tool discovery to confirm the operations actually exposed by your service.

Deploy a workload

1. Open Edges → Workloads and start creating a workload.
2. Enter a Name and a container Image available to the cluster. Prefer a fixed version or digest for repeatable deployment.
3. Choose Replicas and Strategy: Spread targets all matching edges; Singleton selects one.
4. Enter an Edge selector using comma-separated labels such as env=dev. Confirm those labels belong only to the Kubernetes edges you intend to use.
5. Choose Create workload. Inspect the resulting Placements to see which edges received the workload.
6. Wait for ready replicas, then check the application’s behavior on the target cluster. A running container does not necessarily expose a public URL.

For a marketplace application, select the application, enter Workload name, choose a Kubernetes Edge, review the pinned chart and service information, then choose Deploy. This path creates a singleton Helm workload pinned to that edge; complete any service credentials afterward.

If the service or workload is not ready

Optional CLI diagnostics

If the console status does not explain a failure, select the owning Faros workspace , then list deployment and placement state:

kubectl faros get workloads
kubectl faros get placements
kubectl faros edge list

Workloads report image, phase, and ready replicas. Placements report the target edge, phase, and ready replicas. Verify that placement targets match your intended edge labels. No Placements suggests selection or reconciliation needs investigation; disconnected edges explain a different class of failure. A ready placement still needs an application-level check on its target.

Cleanup

Delete the test Workload and confirm that its Placements and target resources are cleaned up. Review chart persistence settings before assuming stored data is removed. Remove a test Service when its MCP exposure is no longer needed.

For target connection and credentials, see the edge quickstart .