Workspace
A workspace binds a repository to a chat. You say what you want; the agent asks what it still needs to know, then starts working.
Interlaken Cloud
A prompt becomes a live application. Underneath it: real machines, private networks, disks and databases.
A prompt becomes a live application. Underneath it: real virtual machines, private networks, replicated disks and managed databases, ready whenever you want to go deeper.
01 From prompt to URL
An animated illustration: a prompt is typed into Interlaken, which plans the app, provisions a PostgreSQL database, builds a container image, deploys it behind a load balancer, issues a TLS certificate and opens the finished page in a browser window at https://barbershop.interlaken.ai.
02 Two front doors
Start by saying what you want. Drop down to the machines whenever you need to. Both doors open onto the same infrastructure, the same account and the same bill.
Write a prompt. Interlaken generates the application, provisions the database it needs, builds it and puts it behind a URL.
Go straight at the primitives: machines, private networks, replicated disks, databases and Kubernetes — with an API, an MCP server and role-based access over all of it.
03 Capabilities
Seven areas, one account, one network, one bill.
Git · App Builder · WordPress
Describe an app or push a repository. Interlaken builds it and hands back a URL.
VMs · microVMs · containers
Virtual machines, Firecracker microVMs and containers on one fabric.
VPC · load balancers · egress
Private networks, load balancers, public addresses and routed egress.
LINSTOR · DRBD · MinIO
Replicated block volumes, snapshots, images and S3-compatible buckets.
Five engines · KubeBlocks
Managed PostgreSQL, MySQL, MongoDB, Redis and Elasticsearch.
k3s · CNI · CSI · kubeconfig
Managed clusters on your own machines, with storage and load balancing wired in.
Claude · Codex · MCP
Development agents with their own sandbox, terminal and repository access.
04 Underneath
Every application gets a private network. Traffic arrives at a load balancer, crosses into your VPC, reaches your machines, and lands on disks replicated to more than one host.
Counts are what the platform ships; the microVM boot figure is approximate, measured on our own hosts.
05 AI agents
An agent gets a real sandbox in your own tenant — a machine with a filesystem, a terminal and the repositories you hand it. Give it a role, put it in a team, connect Claude or Codex.
A workspace binds a repository to a chat. You say what you want; the agent asks what it still needs to know, then starts working.
Every run is its own microVM with its own kernel and disk. The agent runs commands, fixes what breaks and never shares a sandbox.
The code stays in your repository the whole time. Every step the agent takes is written into the chat, so nothing happens out of sight.
The agent commits and pushes the branch. The push triggers a build on your deploy server, and the app updates on its own URL.
Start
Create an account, type one sentence, and see what comes back. Billing is usage-based — you pay for what you run.