Anyport Cloud

Anyport works with clusters you bring, and it can also run one for you. A cluster on Anyport Cloud is a cluster like any other in your organization: it appears in the cluster list, holds projects, apps and managed services, and is deployed to the same way. The difference is whose hardware it runs on, and so who operates it and who pays for the compute.

Use it when you have no cluster yet, or want one without looking after a machine. Attach your own whenever you like; both kinds sit side by side in one organization.

Creating one

In the console, open Connect a cluster and choose Run it on Anyport Cloud. From a terminal:

anyport cluster create my-cluster --cloud

--region picks a region; without it, the first one with room is used. Nothing is installed on your machine and there is no token to handle: Anyport places the cluster on one of its hosts, installs the agent inside it, and the cluster shows as connected once it is ready. You can create projects and apps straight away. Anything you create waits for the cluster and deploys when it is up.

If a region is full, creation is refused rather than queued somewhere else. Try another region, or try later.

What it costs

Anyport Cloud runs on prepaid credit. There is no free compute, so the first cluster needs a credit purchase (or a Pro plan) before it can be created. Each day's usage is rated and drawn from your balance: CPU and memory for what your apps use rather than what they reserve, storage at the size a volume is provisioned, and builds by the minute. The cluster's own control plane is billed the same way, so even an idle cluster draws a little. There is no per-cluster base fee.

Rates, packs and what Pro includes are on the pricing page. Your balance, how long it will last at the current rate, and each cluster's usage are under Settings → Usage in the console.

When credit runs low you are emailed with about a week left. At zero, what is running keeps running, but new deploys, builds and service installs are refused. After three days at zero the cluster is paused with its data kept, and after 30 days paused it is deleted, with an email before each step. Topping up at any point undoes it.

How large it can be

The cluster page shows how much of that allowance is in use. A deploy or a volume that would go past it is refused, and the refusal names the limit.

Builds have room of their own, apart from your apps: one at a time, sized by Anyport to the machines they run on. A second build waits in the queue until the first finishes.

What works differently

A cluster on Anyport Cloud runs on machines shared with other organizations' clusters, so a few things that are yours to decide on an attached cluster are Anyport's here.

• Anyport operates it. The agent, its upgrades, the ingress controller, HTTPS and build settings are managed for you, and the console does not offer to change them. Upgrades are scheduled by Anyport and announced on the cluster page a day ahead; a routine one can be deferred once by 14 days, a security one cannot. A snapshot is taken first, your apps keep serving, and the cluster's API pauses for about a minute while its control plane restarts.
• HTTPS only. Apps are published over HTTPS on generated URLs and your own domains. Public TCP and UDP ports are not offered. To reach a port from your own machine, use anyport port-forward.
• Memory is what you ask for. An app's memory limit is set to its memory request, so to give an app more memory, raise the request. CPU has no limit beyond the cluster's allowance.
• Volumes may have a minimum size. Where a host provisions volumes with a minimum, the volume field says so, and that minimum is the size billed.
• Logs and metrics are included. There is no Observability service to install: searchable logs and metrics over time work from the start, with logs kept for 7 days and metrics for 15. A cluster that writes more than 512 MiB of logs in a day has further lines dropped until midnight UTC, with a line in its logs saying so.
• Some services are not offered. GitHub Actions runners need privileged containers, and the Observability service reads node log files; neither is available on Anyport Cloud, and the catalog leaves them out.
• Isolation is Anyport's. Every pod runs in a gVisor sandbox. The cluster is kept apart from every other Anyport Cloud cluster and from the machines it runs on: nothing outside reaches its apps except through their public addresses, and its apps cannot reach other clusters, private networks or outgoing mail ports. Projects inside the cluster can reach each other.

Everything else is the same as on a cluster you attach: builds from GitHub, previews, managed databases with backups to your own bucket, tasks, secrets, shell access, config as code, and duplicating a project onto another cluster.

Deleting one

anyport cluster detach my-cluster

An Anyport Cloud cluster can only be deleted with everything in it: every project, its apps, databases and volumes. The command asks first and waits for the teardown; Anyport then removes the cluster from its host. --keep-agent, --keep-workloads and --purge apply only to clusters you attach.

To keep something, duplicate the project onto another cluster first.

Next steps

• Pricing — Rates, credit packs and what Pro includes.
• Connecting clusters — Attach a cluster of your own instead, or as well.
• Quick start — From sign-in to a running app.