AI
Agent Runtime
Agents request capabilities through policy instead of owning unrestricted shell authority.
Server Edition
Governed intelligence for infrastructure you control.
A headless AI-native system for private clouds, VMs, bare metal, model routing, agent execution and controlled automation—without turning the model into root.
Headless by design
Small image. Strong trust boundaries.
The Server edition focuses on the agent runtime, model fabric, memory, policy, audit, system health and update chain rather than carrying a desktop environment.
✓Unix-socket control plane▣Capability-gated execution↺A/B update architecture
M
Model Fabric
Route workloads across local and cloud providers by privacy, capability, latency and availability.
P
Policy & Audit
Identity, approvals, capability checks and audit remain independent from any model.
H
Health Intelligence
Structured health state helps operators see update posture, risk and recommended next actions.
Update and recovery
Designed to move forward without losing a known-good system.
ROOT_A / ROOT_B, encrypted STATE, health confirmation and recovery are treated as one operating lifecycle.
Run AKnown-good active slot
Stage BWrite inactive slot
Health checkConfirm after successful boot
RollbackReturn to confirmed slot on failure
Server Developer Preview is available now.
Use it for evaluation, VM testing and development—not as a production Stable release.