Make agent sessions, memory and decisions observable, bounded and reviewable.

SONUV does not replace agent planning or human judgment. It carries session context, exposes execution boundaries and leaves a decision trail for review.

Private prototype / Shadow mode
Session demoEvidence & boundary

Capabilities

Three capabilities form the minimum governance loop.

01

Session Governance

Connect session start, stop and behavior observations so the executor, task and decision can be reviewed together.

02

Memory Context

Provide bounded relevant context at task start instead of repeatedly loading an entire conversation history.

03

Decision & Audit Trail

Separate proposal, authorization, execution and review, keeping a human gate for consequential actions.

Governed session demo

Enter a task to see how SONUV expresses a governance observation. This demo runs only in the browser and calls no agent, model or backend.

simulation: true
No input is sent and no action is executed.
Governance observationidle
{
  "simulation": true,
  "session_state": "awaiting_scenario",
  "external_effect": false
}

What current evidence supports

The public page describes mechanisms present in the local runtime without presenting the internal prototype as a public platform.

  • A single session CLI records start, stop, status and review.
  • Multiple local agent clients connect to the session lifecycle.
  • Shadow mode records governance recommendations, behavior observations and agreement analysis.
  • Memory context and governance logs are separated rather than injecting all history into each session.

Current product boundary

  • No public signup, multi-tenant SaaS, production SLA or open commercial API.
  • The page demo is simulation: true; it sends no input and triggers no real execution.
  • AIOBR is a protocol and trajectory-expression layer, not an agent delegated by SONUV.
  • SONUV governs plan and execution boundaries; a human still approves consequential actions.