Agent Service NowTurn AI into governed,
private productivity for your enterprise
One control plane that meters, protects and governs AI on your private compute — for Trio clients and for the agents you already run.
Files stay on the device. Tools run in an OS-native sandbox; OCR and PII checks run locally before anything is sent.
Not using Trio? Your own agents connect through the gateway with one key per agent.
How it works
Apps on top are your choice. The compute underneath is always private. ASN sits between them, so every model call passes one set of rules.
Desktop, browser, scheduled tasks, or any app that speaks the OpenAI-compatible API.
Checks who is calling, masks personal data, applies model and content rules, reserves credit, and records usage without storing content.
Cost per token is part of the bill. Mixed hardware underneath; one meter on top.
See the gateway mask a prompt
Sample text only. In reversible mode the model sees placeholders and the answer comes back with the original values restored.
Please check last month's statement for client Alex Chan (HKID A123456(7), account 012-345-67890, email alex.chan@example.com).
Please check last month's statement for client <PERSON_1> (HKID <HKID_1>, account <ACCOUNT_1>, email <EMAIL_1>).
Placeholders are swapped back to the real values in the reply the employee sees.
The deeper you connect, the more you can govern
Every deployment starts with tokens and cost. Each level adds switches in the control tower; you pay for the switches you turn on.
One API key against our private compute.
Token usage and cost, model allowlist, budgets and credit pools, credential filter.
One key per agent, or an agent-ID header.
Per-agent usage, PII masking at the gateway, moderation, audit log, department quotas.
Add session and turn IDs plus one completion callback.
Agent-run billing and ROI dashboard (version 2.0).
Run agents in Trio Local, Web or Task.
Agent store and approvals, OS-native sandbox, on-device OCR and PII.
7 deployment profiles
Same control plane, different depths. Pick the layout closest to yours.
Enterprise HQ
A platform that resells AI services to its own members or customers.
Private token factory + ASN + Trio Local, Web and Task
In the platform, with context kept per member
- Different guardrail profiles per audience
- Third-party model calls recorded against each member
- Agent store with approvals for every team
Governance modules
Each module shows the integration level it needs.
Reversible pseudonyms or irreversible placeholders at the gateway; Trio clients also mask on the device.
Personal data reaches the model only as placeholders.
Organization and sub-tenant credit pools with hard stops, approved overage and one-click top-ups.
Spend stays inside the budget you set.
Model allowlist per tenant, credential filter, sensitive-word and content moderation.
One set of rules for every app and every model.
Who called which model through which agent, recorded without prompts or files.
Answers the auditor without creating a new data store.
Counts completed runs and accepted outputs; customer-set baselines turn them into hours and value.
Pay for work done, not tokens burned.
OS-native sandbox for agent tools, on-device OCR and PII masking in Trio clients.
Sensitive files never leave the machine.
ROI you can audit 2.0
Tokens are not productivity. With version 2.0, ASN counts completed agent runs and whether their output was accepted. You set the manual baseline, so the result is your number, not ours.
Govern AI on your own compute
Start with tokens and cost. Turn on more when you are ready.