Private beta · prepaid EU cloud

Run AI code in real microVMs.Not containers.

Sistemo boots a Firecracker microVM for every sandbox — own kernel, isolated network, ready in about a second. One API call, prepaid so you never get a surprise bill.

Private beta. The cloud isn't open yet — everything below is what it will do. Request an invite and we'll email you when it opens.

  • No card to start
  • Python + TypeScript SDK
  • Pause at €0
sandbox.py
# pip install sistemo
from sistemo import Sandbox

sb = Sandbox()          # real microVM
r = sb.run("python3 agent.py")
print(r.stdout, r.exit_code)

Containers share a kernel. That's the problem.

When an LLM writes the code, shared-kernel isolation is not enough. MicroVMs give each run its own kernel — the boundary you actually need.

Typical container sandbox
  • Shared host kernel
  • Escape = host compromise
  • Noisy neighbors on the same node
  • Hard to reason about multi-tenant risk
Sistemo microVM
  • Own kernel per sandbox (Firecracker)
  • Jailer + network isolation on the host
  • Dedicated resources, clean teardown
  • Built for AI-generated & untrusted code

Hardware-level isolation

Every sandbox is a Firecracker microVM. Not a container with extra flags.

One API. Minutes to first run.

Mint a key, Sandbox().run(…). Python and TypeScript SDKs, zero deps.

Prepaid. No surprise bills.

€5 free, no card. Usage draws the balance. Machines pause at €0 — your spend cap is the wallet.

From signup to running code

Three steps. No infra to manage.

  1. 1

    Get an API key

    Open the dashboard, mint a full-scope key. Copy sk_live_… once.

  2. 2

    Install the SDK

    pip install sistemo or npm install @sistemo/sdk — pure stdlib / fetch.

  3. 3

    Run code

    Sandbox().run("…") boots a microVM, returns stdout and exit code.

$ export SISTEMO_API_KEY=sk_live_…
$ python3 -c "from sistemo import Sandbox; print(Sandbox().run('python3 -c \"print(2+2)\"').stdout)"
4
Full quickstart

Built for people who run other people's code

If the code came from a model, a user, or a plugin — give it a machine, not your host.

AI agent builders

Your agent writes code. Run it in a microVM, not next to production secrets.

AI coding tools

Execute model output per request behind one API call. Tear down when done.

Untrusted workloads

User scripts, plugins, eval harnesses — isolation you can explain to security.

Simple pricing

No subscription. Your balance is the spend cap.

Full pricing
Prepaid wallet
€5 free· no card

Add a card (no charge) → €30 total trial

  • Pay per second compute · per GB-month storage
  • Trial: 1 machine · 1 vCPU · 1 GB · 20 GB storage
  • At €0 machines pause — never a surprise invoice
Claim your €5 at launch

Be first on a real microVM

€5 of credit and an API key waiting the day the cloud opens. No card, no commitment.