ttl64
K3s · BR & US · now in beta

Kubernetes sandboxes in seconds.

A real K3s cluster on microVMs, with a public endpoint and kubeconfig. Use it, let it expire. You pay only for the computers while they run.

$ curl -fsSL https://ttl64.computer/install | sh
~/projects/apizsh
why the name

Every packet expires.
Your clusters should too.

Linux sends every IP packet with a time-to-live of 64. When it hits zero, the packet is gone — that’s why the internet doesn’t fill up with lost traffic. ttl64 applies the same rule to Kubernetes.

64
ttl · 64 triangles, one per hop

Create. Use. Expire.

$ ttl64 create --ttl 1h

Create

A control plane and worker nodes, each its own microVM, with a public endpoint and kubeconfig. Ready in about 20 seconds, from the browser or the terminal.

$ kubectl apply -f .

Use

Real Kubernetes, not a simulator. Deploy, test, break things. Need more time? Extend the TTL with one command.

# ttl 0 → gone

Expire

When the timer hits zero, the cluster and everything in it is destroyed. Nothing to clean up, nothing left running.

scale

Computers when you need them

Point at a height on the rack. Computers drop out of the cloud into their slots, one after another.

expire

Gone when the TTL ends

Slide along the hours. In the last one the nodes power down, the control plane last.

Built for throwaway work

Things you shouldn’t keep a cluster running for.
Pull request environments
One cluster per PR, gone when the review is done.
ttl 4h
CI integration tests
Real API server, real CRDs, no kind-in-docker gymnastics.
ttl 30m
Trying a Helm chart
Install it somewhere that isn’t your laptop or staging.
ttl 1h
Workshops and training
A clean cluster per attendee that cleans itself up.
ttl 8h
AI agents
Give an agent a cluster it can’t keep forever.
ttl 30m

You pay for the computers
running your cluster.

A cluster is a control plane plus worker nodes, and each one is a computer. Clusters are free; computers cost by the hour. Every started hour is charged in full.

Small
1 vCPU · 2 GB
R$ 0,69 / computer-hour
Medium
2 vCPU · 4 GB
R$ 1,29 / computer-hour
Large
4 vCPU · 8 GB
R$ 2,39 / computer-hour
control-plane Small + node-1 Small2 × R$ 0,69 = R$ 1,38 / hour · TTL 4h → at most R$ 5,52
Load balancer
Public address for a service
R$ 30,00 / month
≈ R$ 0,04 / hour
Storage
Volumes, by the size requested
R$ 1,00 / GB-month
≈ R$ 0,0014 / GB-hour
Developer
R$ 49 / month

Multi-node K3s on Small computers.

  • 50 computer-hours included
  • Up to 2 computers per cluster
  • Small computers
  • 2 clusters at a time
  • TTL up to 8 hours
  • 1 region
Pro
R$ 129 / month

Bigger clusters, every size and region, longer TTLs.

  • 160 computer-hours included
  • Up to 4 computers per cluster
  • Small, Medium and Large
  • 5 clusters at a time
  • TTL up to 24 hours
  • Every region: BR and US
  • Early access to new features

Plans are paid up front each month, by card or Pix. A Small computer uses 1 included computer-hour per hour, Medium 2, Large 4; beyond them, each hour is charged at the size's price. Load balancers and storage are billed per started hour (the monthly price ÷ 730) and never use included computer-hours. Prices in BRL.

Your cluster is
30 seconds away.

$curl -fsSL https://ttl64.computer/install | sh
or use the web app