MAQPNA

Compare

Containers vs sandboxes for AI agents.

Agents run code that a model wrote a second ago and nobody reviewed. Which AI agent sandbox is enough for that?

Side by side

Standard container (runc)gVisor · tier-0Kata + Firecracker · tier-1Confidential VM · tier-2
Kernel boundaryShares the host kernelA user-space kernel intercepts system callsIts own guest kernel in a hardware-virtualised microVMIts own guest kernel; memory encrypted from the host
Protects from the host operatorNoNoNoYes, within the hardware vendor's trust model
System-call compatibilityFullMost workloads; some system calls are not implementedFullFull
Start-up and overheadLowestLowHigher: a VM boots per sandboxHighest: VM boot plus attestation
Node requirementsNoneNone beyond the runtimeHardware virtualisation (KVM)AMD SEV-SNP or Intel TDX hardware
Good forCode you wrote and reviewedCode interpreters, data analysis, untrusted scriptsBuild tools, browsers, broader system-call needsRegulated data, customer secrets, untrusted infrastructure

Where MAQPNA fits

Every tier, chosen per agent, governed the same way.

agent.yaml
apiVersion: maqpna.com/v1alpha1
kind: Agent
metadata:
  name: code-interpreter
spec:
  image: registry.example.eu/agents/interpreter:1.2
  tier: tier-0            # gVisor; tier-1 microVM; tier-2 confidential VM
  policyRef: interpreter-tools

FAQ

Common questions

Is a container enough to run AI agent code?

For code you wrote and reviewed, usually yes.

Do I need confidential VMs?

Only when the host operator is outside your trust boundary, for example on shared or third-party infrastructure, or for regulated data that must be protected in use.

Can I mix tiers in one cluster?

Yes.

Pick the boundary per agent.

Run tier-0 on a laptop today, then choose microVMs or confidential VMs per agent in your cluster.

More install options
curl -fsSL https://maqpna.com/install.sh | sh