Product
One console where agents and humans run the business.
Every screen below is the live product — the same ontology, the same authorization table, the same causal log.
Command center
Supervise by exception.
KPIs, risk signals and agent plans on one surface. A replenishment plan arrives with its reasoning — approve it, veto it, or trace every step back to the signal that started it.
- Agent plans with one-click approve / veto
- signal → decision → action → effect, fully traceable
- Risk surfaced before it becomes cost
Authorization
One permission table for humans and agents.
Every guarded action carries its human roles and the agent’s autonomy tier. Registering a new action is data, not a deploy — add it on this page, gate the code path with one line.
- Autonomy L0–L4 per action, capped by governance locks
- Human roles and agent tiers on the same row
- Add, edit and remove actions right from the matrix
Governance
Even the navigation is an ontology.
The menu tree is data: rename, regroup, disable — and the sidebar, the command palette and the permission matrix all read the same tree.
- One tree drives the sidebar, ⌘K and permissions
- Move sections across groups and planes
- Factory rows protected; every change causally logged
These screenshots are the actual product — Chinese-first UI, bilingual by design.
Run your business with agents.
We are onboarding a small group of early design partners across global commerce.