FuseGov Reference Architecture
A complete, production-grade architecture for governing agentic AI. Read in order or jump to specific topics.
Recommended Reading Order
Hybrid Runtime Governance for Agentic AI
The foundation: why governance must become runtime infrastructure, and how the hybrid gateway + sidecar + evidence pipeline architecture delivers it.
Gateway/Sidecar + Evidence Pipeline
Deep dive into the hybrid enforcement pattern: when to use gateway, when to use sidecars, and how the evidence pipeline ties it together.
Hybrid Runtime Governance (Detailed)
Comprehensive walkthrough of two-stage decisioning (deterministic + semantic), approvals, degraded mode, and pilot rollout strategy.
The Tool Registry: Governing the Action Surface
How to inventory and govern the tools agents can use. Risk tiers, scope constraints, rate limits, and approval rules.
Policy Bundles: Versioned, Signed Controls
Why policies must ship like software. Bundle structure, signing, drift detection, and rollback capabilities.
The Operational Authenticity Layer (Capstone)
CapstoneThe complete reference architecture: policy lifecycle, registries, hybrid enforcement, approvals, evidence, drift, testing, and operating metrics.
Golden Path Demo: Intent → Action → Outcome
Guided walkthrough of the closed-loop: an agent attempts a high-risk action, enforcement escalates to a human, and outcomes are verified.