Managed Kubernetes, Compute & API

You can now deploy and run production Kubernetes clusters directly from the dashboard. No control-plane setup, no rewriting kubeconfigs to get started. Pick your nodes and deploy.

Free control plane, pay only for what runs

The control plane is free. You pay for your worker nodes, billed exactly like any other compute, with no premium. If you need resilience, you can add a high-availability control plane for a flat $30/mo. Your monthly cost is shown live as you build the cluster, so there are no surprises.

Guided create wizard

A five-step wizard walks you through everything: cluster name and Kubernetes version, networking and service CIDRs, your worker node plan, optional storage nodes, and dedicated resources. Then a summary and cost roll-up so you always know the size and price of what you are about to deploy before you commit.

Node pools, labels, taints, and autoscaling

Organise your workloads with node pools. Set per-pool minimum and maximum scaling, attach labels and taints to control scheduling, and turn on automatic scaling. When pods cannot be scheduled because the cluster is short on CPU or memory, Raff adds worker nodes up to your maximum. When usage stays low for a sustained period, it scales back down to your minimum. Scale-up reacts to pods that are already waiting on capacity, so expect a short reaction window of a few minutes rather than instant or predictive scaling.

Dedicated storage nodes for persistent volumes

Need persistent storage for stateful workloads? Add up to three dedicated storage nodes, each sized 100 to 1000 GB, and Longhorn is enabled automatically. Storage capacity is provisioned to the size you choose, and the cost is folded into your cluster's monthly total. Clusters without storage nodes simply skip Longhorn, so there is no half-configured storage layer sitting idle.

A real detail page for every cluster

Each cluster gets a full detail view with dedicated tabs.

• Overview: cluster health, KPIs, and a live events timeline that streams provisioning progress from 5% to 100%.
• Nodes: every node with its role and health status.
• Pools: manage your node pools.
• Workloads: see what is running across your namespaces.
• Metrics: live CPU and memory for the cluster and each node, refreshed automatically.
• Logs: a built-in log terminal with pause and download.
• Settings: rename the cluster, manage the firewall, and a danger zone for tear-down.

If a cluster ever becomes unreachable, the dashboard tells you clearly instead of showing stale state. Every cluster also ships with a curated secure baseline firewall that actually enforces. See the fixes on this line for why that matters.

We rebuilt the compute list and every VM detail tab to match our new design language. It is cleaner, faster, and shows you more of what matters at a glance.

Live CPU and RAM on the list

Your VM list now shows live CPU% and RAM% straight from each VM, updating in real time, with no page refresh and no polling. New VMs show their specs immediately, and the live bars fill in within about a minute as the first stats arrive. High-CPU or high-memory VMs are flagged in red so problems jump out.

Redesigned Resources list

The compute list has a new header with a live summary line (for example 32 VMs · 8 running · 3 regions · 128 GB) that adapts to whichever tab you are on. New filter chips let you narrow to running VMs or a specific region, and a three-way toggle lets you group your fleet by region or keep a flat list. Bulk power actions live in a slide-in action bar. Cards are tidier: abbreviated status pills, a monospaced spec strip, and clean three-line resource readouts for vCPU, RAM, and disk. Everything adapts down to tablet and mobile layouts.

Every VM detail tab, polished

• Information is a two-column Notes card (personal and team notes, with "edited by · time ago").
• Metrics is a 2×2 chart grid with five time ranges (1h, 6h, 24h, 7d, 30d) and clearer network and disk read/write series.
• Networking shows separate IPv4, IPv6, and VPC cards, with clear reboot warnings when you change firewall assignments.
• Snapshots & Backups are redesigned single-VM views with a clear free-pool quota strip.

Resize and Actions, refreshed

The Resize tab now shows a side-by-side current vs. new comparison with an exact charge or credit preview before you confirm, and a "Stop VM first" button right where you need it, with a live "Stopping…" state. The Actions tab gets a graceful Reboot option alongside the existing hard reboot, and we unified the power-cycle naming everywhere so it is consistent across VMs: Reboot / Restart for a graceful cycle, Hard reboot / Hard restart for an unresponsive VM.

Faster, flicker-free loading

We reworked loading states across compute and VM detail so you see a single skeleton that matches what is about to render. No more generic placeholder flashing, then a spinner, then the real page.

We replaced the old "first 3 backups per VM are free" rule with something simpler and more generous: one shared free-storage pool per account that covers both backups and snapshots.

How it works

Every account gets 30 GB free to start, and the pool grows with your fleet: 10% of each standard VM's storage is added to your free pool (15% for premium VMs). You only pay for storage beyond the pool, the same way as object storage overage. A new usage card on both the Backups and Snapshots tabs shows exactly how much of your pool you are using, and your VM plan card shows how much free storage each VM contributes.

Multiple schedules per VM

You are no longer limited to a single backup schedule per VM. Add as many as you need (for example a daily and a weekly) and they share a common retention setting for that VM.

Retention is now "keep N most recent"

Retention always counted the number of backups to keep, but the UI used to describe it confusingly in days and weeks. It now says exactly what it does: "keep the N most recent", and that count applies to every backup of the VM, including manual ones. Weekly schedules now show a real countdown to the next run, not just a date.

Backups and snapshots, side by side

Our incremental backups use a storage mechanism that cannot run at the same time as disk snapshots on the same VM. Rather than queueing, we made the rule one direction and self-healing: while a VM is using backups, taking a new disk snapshot is blocked, and the Snapshots tab explains why with a one-click link over to Backups. The moment you remove the last backup, snapshots work again automatically. Backups are never blocked by snapshots.

Browser terminals used to drop into a permanent error after 20 to 30 minutes and lose your shell. Not anymore. Your shell now persists across disconnects, and the browser automatically re-attaches to the same session with a smart reconnection. Kill the connection, reopen, or go idle for half an hour, and your long-running commands keep running so you pick up right where you left off. Closing the session still ends it cleanly.

The cloud firewall is now a first-class part of the VM Security tab instead of a read-only table tucked at the bottom. It sits at the top so it is your primary surface, with one card per public network interface. You can assign, edit, and delete firewall rules right from VM detail, and an unconnected interface is clearly flagged with an "Open to internet" warning and a one-click way to lock it down. Rule changes are enforced immediately without a page refresh, and the edit dialog reminds you that rule changes take effect after a full power cycle.

Cancelled subscriptions disappear cleanly

Fully cancelled subscriptions are now hidden from your billing view right away, instead of lingering for a month. Subscriptions set to not renew still show their "Won't renew · Ends [date]" state, and you can still resume them until the period actually ends.

Kubernetes shows up properly

Kubernetes clusters now appear in your billing and subscriptions with their real cluster name and a readable "Kubernetes" type, just like VMs and volumes, instead of a raw internal label.

You can now build on Raff three ways: a REST API, a command-line tool, and a Terraform provider. All three talk to the same public API, use the same API key authentication, and respect the same role-based scoping, so a key you create once works everywhere.

Public REST API (Beta)

The Raff public API is live in beta. Authenticate with an X-API-Key header against https://api.rafftechnologies.com and manage compute, storage, networking, and your team and access controls programmatically. The endpoint surface is stable, with an interactive playground in the docs to try calls before you write code. Generate a key from the dashboard under Team & Projects, then API Keys.

Raff CLI

raff is a single static binary for macOS, Linux, and Windows that wraps the full public API for the terminal. It is built for CRUD-style operations, scripts, and CI/CD pipelines, and it covers the full lifecycle for VMs, volumes, snapshots, backups, VPCs, public IPs, firewalls, SSH keys, projects, and IAM. Install it with go install github.com/RaffTechnologies/raff-cli/cmd/raff@latest or a release binary, run raff configure, and create a VM with raff vm create --name web-01 --template 12 --plan small --region us-east. Every command supports --output json for piping into your own tooling, and named profiles let you keep prod, staging, and dev credentials side by side.

Terraform Provider

For infrastructure as code, the Raff Terraform provider is published on the Terraform Registry as RaffTechnologies/raff, with 14 resources covering compute, storage, networking, and IAM, plus 25+ data sources for discovery and pricing. Version your VM, project, and networking definitions in git, review them in pull requests, and apply them with terraform apply.

The CLI is imperative and the provider is declarative, so use whichever fits the job. Kubernetes and Object Storage are still rolling out at the API level and will reach the CLI and Terraform once those endpoints stabilize.

Every fix below is tagged with the patch it shipped in, so you can tell what is already on your account.

• Kubernetes firewall now actually enforces. Cluster firewalls were applied but not blocking: a default "allow all" rule was silently overriding the secure baseline. Shipped in 1.5.1.
• Incremental backup chains no longer over-delete. Cleanup could delete the whole chain, including newer points that were still valid. Shipped in 1.5.1.
• Snapshot rename and delete errors. Renaming or deleting a snapshot could fail with an internal error. Shipped in 1.5.1.
• Long VM names no longer break dialogs. Create Snapshot and Backup/Snapshot dialogs overflowed for very long names. Shipped in 1.5.1.
• Reinstall OS keeps your public IP. Reinstalling a VM's OS no longer detaches and re-attaches the public IP. Shipped in 1.5.2.
• Reinstall OS keeps your IPv6. The IPv6 address is preserved through the reinstall. Shipped in 1.5.2.
• Pro-rata refunds are accurate. Refunds now credit to the hour instead of rounding to the start of the day. Shipped in 1.5.2.
• Account deletion is safe. Deleting an account with an active subscription now tells you exactly what to clean up first. Shipped in 1.5.2.
• Email changes propagate everywhere. Invoices and payment emails follow the new address. Shipped in 1.5.2.