Skip to content

Platform

Everything a cloud does. One console.

Machines, networks, disks, databases, clusters, deploys and agents — created through one API and one console. Underneath sits one spine: tenants, roles, quotas, and every resource metered into a single invoice.

Open the console See the infrastructure

01 / 07 Deploy

From repository to address

A managed application platform, built on Coolify and run for you. Push a repository or describe the app — either way you end up with a build, a running process and a URL.

Deploy servers
A managed Coolify instance of your own, provisioned and kept running by the platform.
Deploy from Git
Point at a repository and pick a runtime. Interlaken clones it, builds it and runs it.
Config variables
Edit environment variables in the console and apply them to the running app.
Attached databases
Bind a managed database to an app; the platform wires the connection details in.
App Builder
Describe an app, watch agents build it against a repository, get a live URL — and roll a bad deploy back.
WordPress sites
Managed WordPress, provisioned and operated like any other application.
bakery — deploy healthy

git push interlaken main

  1. runtime detected — node 22
  2. build — npm ci && npm run build
  3. image 9f3c1a2 pushed
  4. rollout — 1/1 healthy

live at https://bakery.apps.interlaken.ai

02 / 07 Compute

Run it however it wants

Four ways to run a workload, all created from the same console: a whole machine you log into, a microVM you start for one job, a container that is only its image, or a pool that sizes itself. Whichever you pick, it lands on your networks and your disks.

Virtual machines
Launched from cloud images with cloud-init, so your SSH key and first-boot configuration are already in place.
microVMs
Built on Firecracker, one jailer per guest, launched from a rootfs image you prepared.
Containers
OCI images on the same networks and volumes as every other workload.
Autoscaling groups
Grow and shrink a pool of machines or containers against CPU and memory metrics.
Hot attach and detach
Add or remove disks and network interfaces on a running virtual machine.
Consoles
VNC for graphical machines, serial for microVMs — in the browser, without a bastion host.
  • 01

    Virtual machine

    A whole operating system, brought up with cloud-init and yours to log into.

  • 02

    microVM

    A Firecracker guest you start for one job and let go of afterwards.

  • 03

    Container

    An OCI image, run as it is, on the same network as everything else.

  • 04

    Autoscaling group

    A pool of any of the above, grown and shrunk against a metric.

Shared: networks · disks · images · metrics · console

03 / 07 Networking

A network you actually configure

Every tenant gets isolated virtual private clouds with their own address space. Route out through a gateway, put a load balancer in front, and hand out public addresses from a managed pool.

VPCs and subnets
Isolated address space per tenant, segmented into the subnets you define.
Gateways and public IPs
1:1 NAT egress per VPC, with public addresses allocated from a managed pool and bound to an interface.
Load balancers
Layer 4 pass-through, TLS termination, HTTP routing and HTTP/3 over QUIC — with health checks.
VPC peering
Connect two private networks to each other without going out to the internet.
DHCP and DNS
Option sets you attach to VPCs and subnets, and per-VPC records served by the gateway.
Bandwidth limits
A rate cap per network interface, enforced in the data path rather than inside the guest.
vpc-main · ingress health-checked
Internet gateway · 1:1 NAT 203.0.113.10 load balancer ‎:443 vpc-main · 10.0.0.0/16 dhcp · dns subnet-a · 10.0.1.0/24 web-01 10.0.1.11:8080 web-02 10.0.1.12:8080 peered VPC rate cap

04 / 07 Storage

Volumes, never one machine’s problem

Block storage is replicated across hosts with LINSTOR and DRBD, so a volume lives on more than one machine from the moment it exists. Snapshots become bootable images; unstructured data goes into S3-compatible buckets alongside.

Replicated block volumes
Each volume is placed on several hosts, and every write is mirrored to the replicas.
Snapshots
Point-in-time copies of a disk that you can restore into a new volume.
Golden images
Publish a snapshot as a bootable image and launch new machines from it.
Object storage
S3-compatible buckets on MinIO, with access keys scoped to a single bucket and a file manager in the console.
Online expansion
Grow a volume and its filesystem while the workload using it keeps running.
vol-8f2a1c · 100 GiB in sync
host-01 primary
host-02 replica
host-03 replica
write One volume, more than one machine. Example volume.

05 / 07 Databases

Managed engines, replicated storage

Five database engines, provisioned and continuously reconciled by KubeBlocks on the platform’s own Kubernetes, with their data on the same replicated block volumes as everything else. You get an endpoint and credentials.

Five engines
PostgreSQL, MySQL, MongoDB, Redis and Elasticsearch.
Operated, not just installed
Clusters are provisioned and continuously reconciled — you never write a manifest.
Replicated storage
Database volumes are the platform’s replicated block volumes, with the same snapshot support.
Private by default
Reachable from inside your VPC; exposure beyond it is something you ask for.
Attachable
Bind a database to a deployed application and skip copying credentials around by hand.
Engines KubeBlocks · reconciled
  1. 01 PostgreSQL Relational
  2. 02 MySQL Relational
  3. 03 MongoDB Document
  4. 04 Redis Key-value
  5. 05 Elasticsearch Search
replicated block volume

endpoint + credentials

06 / 07 Kubernetes

A real cluster on machines you see

Managed k3s clusters whose nodes are your own virtual machines, in your own VPC. Networking, the storage driver and load balancing are wired in before the kubeconfig is handed over.

your VPC
kubectl

kubectl get nodes

NAME STATUS ROLES AGE VERSION

cluster-1-cp-1 Ready control-plane,master 12d v1.31.4+k3s1

cluster-1-w-1 Ready <none> 12d v1.31.4+k3s1

cluster-1-w-2 Ready <none> 12d v1.31.4+k3s1

  • CNI
  • CSI
  • LB
Managed k3s clusters
A control plane and a worker pool, provisioned as virtual machines inside your network.
Wired in on arrival
Platform CNI, CSI and load balancing are configured as part of provisioning, not afterwards.
CSI volumes
Persistent volumes on replicated block storage, expandable while pods keep using them.
Kubeconfig download
Take the kubeconfig from the console and point kubectl straight at the cluster.
Addons kept in sync
Cluster addons are reconciled from the platform’s own manifests, so drift is corrected for you.

07 / 07 Agents

Agents with somewhere to work

An agent on Interlaken gets a real sandbox — a machine with a filesystem, a terminal and network access — rather than a chat window. Give it a role, skills and tools, and let a team of them work through one task together.

Sandboxed runtimes
A run is a pausable, resumable fork of a sealed environment you built and versioned yourself.
Teams and roles
A lead delegates to role-assigned agents, every message is recorded, and you can talk to the lead while it works.
Skills and tools
Capabilities are granted explicitly. An agent has what you gave it and nothing else.
Claude and Codex connectors
Store your own CLI credentials per tenant so agents run under accounts you control.
MCP server and OAuth 2.1
Interlaken exposes its own API as MCP tools behind an OAuth 2.1 identity provider, so an outside agent can drive the platform under your roles.
run · fix-checkout-total sandbox · vm-xsmall
  1. plan Locate the failing test and the handler behind it
  2. read api/checkout — 3 files
  3. edit 2 files changed
  4. shell go test ./... — 41 passed
  5. done Pushed branch fix/checkout-total

08 Operations

Operated as one platform

Every resource on this page belongs to a tenant, is created by a user holding a role, counts against a quota and is metered into an invoice. That spine is what makes this a cloud rather than a pile of services.

01 / 06

Tenants

Isolation between organizations, across networks, storage, identity and billing.

02 / 06

Users and roles

Role-based access control over what each user may do with each kind of resource.

03 / 06

Quotas

Per-tenant limits, checked before anything is provisioned rather than after the bill arrives.

04 / 06

Usage metering

Running resources are metered continuously and accumulated into usage records.

05 / 06

One invoice

Usage becomes an invoice. Credits, payment methods and past invoices all live in the console.

06 / 06

Suspension controls

A tenant can be suspended, which cuts its network access first and asks questions afterwards.

Console

See it for yourself

Every section on this page is a page in the console. Create an account and the whole platform is one click away.