← Learn

Sentinel SCA Architecture

Sentinel SCA is a governance runtime for autonomous systems and AI agents. Its purpose is not to make decisions for agents, but to determine whether a proposed action remains admissible before execution occurs.

Core Principle

Proposal
   ↓
Admissibility Evaluation
   ↓
Decision
   ↓
Execution Boundary
   ↓
Receipt
   ↓
Audit & Replay

Sentinel governs execution. It does not replace the agent. It determines whether execution is admissible.

Governance Layers

Identity Verification

Capability Validation

Schema Validation

Risk Evaluation

Policy Evaluation

Decision Generation

Human Review Routing

Execution Window Validation

Execution Boundary Enforcement

Receipt Verification

Audit Chain Recording

Replayability

Fail-Closed Governance

Protocol Safeguards

Policy Weapon A: No execution without admissibility.

Policy Weapon B: No admissibility without accountability.

Execution Boundary

The execution boundary is the point where a proposed action becomes real. Sentinel is designed around this boundary.

Execution boundaries may include:

Cloud infrastructure
APIs
Industrial systems
IoT devices
Robotics
Autonomous farm systems
External enforcement services

Receipts and Replay

Every governance decision can be associated with receipts, audit evidence, and replayable records. This enables forensic analysis and deterministic review.

Fail-Closed Design

If Sentinel cannot establish admissibility, the safe outcome is non-execution.

Governance before execution.