Every agent needs a computer.
Raft gives each AI agent its own Linux VM with SSH, Docker and a desktop, from $0.03 an hour. We're not live yet: join the waitlist for early access.
Private beta. Europe opens first, then the US and India.
click the water
to restore a VM from its snapshot
of private memory per restored VM
to hand out a warm sandbox
for 4 vCPU and 8 GB on Raft Cloud
VM numbers are measured on our open-source VMM prototype (Apple M4 Pro, 64 MiB guest). Cloud price is our launch target.
Start a sandbox, copy it, stop it. The cost below moves exactly the way your bill does: per second, and only while a sandbox runs.
Nothing is running, so nothing is billed.
from raft import Raft
hb = Raft() # reads RAFT_API_KEY
box = hb.sandboxes.create(size="4x8", region="eu")
print(box.id, box.ssh)Switch on only what an agent uses. Each add-on is priced per use, so a sandbox without one costs exactly the base rate.
export default { image: "ubuntu-24.04", size: "4x8", region: "eu",}
You pay for the vCPUs and memory a sandbox has, for the seconds it runs. Nothing when it is stopped.
10 × 8h × 30 days × $0.03/h
Others: public list prices for 4 vCPU / 8 GB, checked September 2026, scaled to your size. Raft Cloud prices are launch targets.
The open-source core on your own servers.
- Create, fork, snapshot, stop
- API, SDK and CLI
- Your hardware, your data
- Community support
Plus $0.00125 per GB-hour of memory. Billed by the second.
- 4 vCPU / 8 GB for $0.030 an hour
- No plan minimum, no idle charges
- Regions: EU first, then US and India
- Add-ons priced per use
Raft run for you, on your servers or ours.
- Managed self-host
- SLA and support
- SSO and audit logs
- Volume pricing
| Feature | Raft | Hosted sandbox clouds | Containers you run yourself |
|---|---|---|---|
| Full Linux VM with its own kernel | |||
| Docker and a desktop inside | |||
| Open source, run it on your servers | |||
| Fork a running machine | |||
| Billed per second, nothing when stopped | You pay for servers 24/7 | ||
| Regions | EU, US, India | Mostly US | Wherever you rent |
| 4 vCPU / 8 GB, per hour | $0.030 | $0.33 – $0.76 | Your server bill |
| Setup | One API call | One API call | Weeks of ops work |
“Some providers” means it depends on which one you pick. Prices are public list prices checked September 2026.
Give your agent a computer.
RaftVM isn't live yet. Get on the waitlist and we'll let you in as regions open, Europe first.
One email when your region opens. No spam.