Skip to content
LACE
  • v0.1 Current
  • Python
  • TypeScript Soon

Get Started

Deployment options

LACE runs in four footprints. The governance substrate — permissions, evidence, budgets, approvals, audit — is identical in every one of them.

Footprints

FootprintWhere it runsNotes
Managed cloudOur cloudFastest start; usage-based AI budgets with hard caps.
Private cloudYour VPCSSO/OIDC, custom connectors, BYO models.
On-premisesYour data centerFull platform, your hardware and boundary.
Air-gappedDisconnected enclaveNo external AI calls; accreditation evidence exportable.

Air-gapped deployment

Air-gapped is not a configuration flag over a cloud product. Model serving, embedding, indexing, and orchestration all run inside the boundary on open-weight models you host, so no code path requires an outbound connection to function — including model and platform updates, which are applied from media you control.

Programs with an accreditation requirement should also read LACE-GOV, which packages the platform with BFO/CCO schema packs, cross-enclave federation, and exportable accreditation evidence.

Model choice

Every footprint supports bringing your own models. Managed cloud can use hosted frontier models under budget caps; private cloud, on-premises, and air-gapped deployments serve open-weight models on your hardware. Because enforcement lives outside the model, changing the model does not change what the platform will let it do.