Asset Registry for Intelligent Agents
A reference architecture for classifying, governing, and composing AI assets at enterprise scale.

Reference Implementation · Tutorial · Architecture Docs

ARIA defines a practical operating model for AI assets built on three foundational layers, with additional operational layers for distribution, AI FinOps, and consumption:

• Metamodel: OASF entities, relationships, and lifecycle states.
• Marketplace: GitHub + OCI patterns for publishing and composing reusable assets.
• Governance: Microsoft Purview integration for sensitivity, lineage, and policy enforcement.

Organizations are building agents faster than they can govern them. ARIA provides:

• A shared taxonomy for agents, skills, domains, modules, and records.
• A discoverable marketplace model with versioning and provenance.
• Governance controls that travel with the asset lifecycle.
• A path from architecture principles to working implementations.

• Reference architecture and conference-ready documentation.
• Infrastructure modules for GitHub marketplace and Azure governance.
• A sample C# agent demonstrating OASF governance middleware.
• An ARIA CLI prototype for search, inspect, audit, and install workflows.

• aria: Architecture, docs, tutorial, Terraform, sample agent, and CLI prototype.
• aria-skills: Skills catalog, orchestrator configs, and reusable skill packages.
• aria-gateway: API and UI gateway components for ARIA distribution and access.
• .github: Organization profile, standards, and shared workflows.

The ARIA mark encodes the framework model:

• Hexagonal hub: canonical OASF record.
• Outer ring: registry boundary.
• Four satellite nodes: primary relationship patterns.
• Three colored lines: metamodel, marketplace, and governance layers.

1. Start with the ARIA tutorial.
2. Review the reference architecture.
3. Explore sample agent implementation.
4. Use Terraform modules to bootstrap your own marketplace and governance baseline.

MIT