Agent Enterprise Platform (AEP)

AIX AEP is the enterprise governance and release authority plane for autonomous AI agents. While AIX Console focuses on agent creation and debugging, AEP ensures that agent capabilities are discoverable, access-controlled, cryptographically verifiable, and compliant with enterprise policies.

Why Enterprise AI Needs AEP

Autonomous agents possess tools, access databases, and make operational decisions. Traditional API gateways lack the semantic comprehension required to govern AI agents. AEP introduces four core pillars:


┌────────────────────────────────────────────────────────┐
│                   AEP Governance Pillars               │
├────────────────────────────────────────────────────────┤
│ 1. Capability Catalog: Dual contracts (Provider/User)  │
│ 2. Access Authority: Two-Person Rule & IAM PAM checks  │
│ 3. Application Blueprints: Isolated sandbox runners    │
│ 4. Cryptographic Evidence: SLSA provenance & audit log │
└────────────────────────────────────────────────────────┘

1. Capability Catalog & Consumption Contracts

Before an agent can be consumed in business workflows, its capabilities are registered in the Capability Catalog. Each capability declares its owner, data sensitivity classification, allowed side-effects, and schema invariants.

2. Request & Multi-Person Governance

Access to high-impact capabilities (e.g., executing financial transactions or updating customer databases) requires formal request submission, automated Cedar policy evaluation, and Two-Person Rule (Separation of Duties) approval.

3. Application Templates & Sandbox Execution

Developers build production agent applications on top of verified blueprints (e.g., Fraud Review Desk, Marketing Campaign Copilot). Blueprints enforce network egress isolation and tenant token boundaries.

4. Immutable Audit & Provenance

Every decision, grant issuance, human approval, and agent tool execution is chained into an append-only cryptographic evidence bundle with SLSA provenance receipts.