AI agent authorization | 2026
TrustGate
A runtime authorization gate for AI agents that checks live business evidence before allowing risky actions.
Project snapshot
- Type
- AI agent authorization
- Period
- 2026
- Source
- Google Cloud Rapid Agent Hackathon 2026
Problem
AI agents can propose high-impact actions such as refunds, but production systems need a transparent control layer that decides whether the action is allowed, needs approval, or should be blocked.
Outcomes
Separated model proposals from final business authorization
Made risky agent actions explainable and auditable
Connected live data evidence to policy decisions on Google Cloud
What I built
Runtime gate between AI agents and risky actions
ALLOW, APPROVAL_REQUIRED, and BLOCK decisions
Deterministic policy engine for final authorization
Auditable decision receipts instead of black-box scores
Live evidence ingestion from Fivetran and BigQuery
Data freshness checks against SLAs
Schema compliance checks
MCP endpoint deployed on Google Cloud Run
Small React control-room dashboard
Vertex AI/Gemini integration for agent proposals
Tech stack
Node.jsGoogle Cloud RunVertex AIGeminiBigQueryFivetranMCPReact
Private client work
This project is described as a case study because the client implementation is not published as a public repository.