One control tower, seven depths of use
Every tenant uses our private compute. How many switches they can turn on depends on how deeply their agents are wired in. Trio is not required, but it is the only way to reach every switch.
Agent Service Now
Governance and control plane
AI Factory
Token factory
Edge AI
Edge hardware platform
Agent Service Now: capability × integration level × deployment profile
Deployment reference only. Availability by integration depth does not indicate that a module is enabled for your organization. 2.0 modules are not available yet.
| Capability | Status | Needs | 01 Enterprise HQ L3 | 02 Localized Branch L3 | 03 Telco Cloud L1 | 04 Multi-Vendor Edge L2 | 05 Professional Firm L3 | 06 Hybrid Cloud L1 | 07 Compute & MaaS Utility L0 |
|---|---|---|---|---|---|---|---|---|---|
| Token usage and cost | Live | L0 | ● | ● | ● | ● | ● | ● | ● |
| Model allowlist | Live | L0 | ● | ● | ● | ● | ● | ● | ● |
| Credit pools and limits | Live | L0 | ● | ● | ● | ● | ● | ● | ● |
| Credential filter | Live | L0 | ● | ● | ● | ● | ● | ● | ● |
| Per-agent usage | Live | L1 | ● | ● | ● | ● | ● | ● | ○ |
| Smart PII substitution at gateway | Live | L1 | ● | ● | ● | ● | ● | ● | ○ |
| Content moderation | Live | L1 | ● | ● | ● | ● | ● | ● | ○ |
| Agent audit log | Live | L1 | ● | ● | ● | ● | ● | ● | ○ |
| Department quotas | Live | L1 | ● | ● | ● | ● | ● | ● | ○ |
| Agent-run billing | 2.0 | L2 | ● | ● | ○ | ● | ● | ○ | ○ |
| ROI dashboard | 2.0 | L2 | ● | ● | ○ | ● | ● | ○ | ○ |
| Blocked-demand log | 2.0 | L2 | ● | ● | ○ | ● | ● | ○ | ○ |
| Agent store and approvals | Live | L3 | ● | ● | ○ | ○ | ● | ○ | ○ |
| Employee skill requests | 2.0 | L3 | ● | ● | ○ | ○ | ● | ○ | ○ |
| OS-native sandbox | Live | L3 | ● | ● | ○ | ○ | ● | ○ | ○ |
| On-device OCR and PII | Live | L3 | ● | ● | ○ | ○ | ● | ○ | ○ |
| Edge reserve mode | 2.0 | L3 | ● | ● | ○ | ○ | ● | ○ | ○ |
| Skill sunsetting | 2.0 | L3 | ● | ● | ○ | ○ | ● | ○ | ○ |
Calls that skip the gateway and go straight to a customer's own model. The boundary is drawn at the gateway, and it is drawn on purpose.
Accepted runs × manual minutes ÷ 60 × hourly cost − token cost − platform fee. We measure runs and acceptance; the customer sets the baseline.
For regulated on-site deployments, the control plane ships inside the appliance. Only counts and module status sync out, never prompts or files.