CROA — Constrained Reachability Orchestration Architecture: A Framework for Deterministic Governance of Agentic AI Execution

CROA (Constrained Reachability Orchestration Architecture) is a vendor-neutral architecture framework for the governed execution of agentic AI systems. It addresses a problem that model-level alignment does not: as AI systems move from advisory tools to operational actors that read, write, and change enterprise systems, how can an organization guarantee what those systems are structurally permitted to do? CROA moves governance from model behavior to the execution layer. Its central claim is stated conditionally and is open to challenge: within a modeled action space, under a registered set of invariants, and given network-enforced execution-boundary containment, registered-invariant-violating execution paths are structurally unreachable — not merely discouraged. Every governance decision, permit and deny, is recorded in an append-only, tamper-evident audit record. This document is the Public Review Draft (v1.0) of the framework. It comprises seven Parts — Foundations, Reference Architecture, Method, Deployment Models, Threat Model, Conformance and Maturity, and Governance — and supporting appendices, including a conformance model (levels L0–L5), machine-readable schemas, sector worked examples, and a minimal reference-implementation pattern. It introduces concepts including execution-layer governance, structural reachability, and Technical Sycophancy.