How do I maintain an inventory of AI models and agents across my company?

Summary

• Use the AI Asset Registry within Unity Gateway, which provides a central inventory for all AI models, agents, MCP servers, tools, and skills built on Unity Catalog.
• Govern all AI assets in one place using the same Unity Catalog access policies you use for data, eliminating fragmented systems.
• Teams publish AI assets once, assign permissions, and make them discoverable and reusable across the organization for users and agents.
• Gain centralized visibility through permissions, lineage, observability, and cost management while Unity Gateway enforces runtime controls.
• All AI assets inherit standard metadata management, carry owner and tag information, and can be discovered alongside other Unity Catalog assets in Catalog Explorer.

How do I maintain an inventory of AI models and agents across my company?

To maintain an inventory of AI models and agents across your company, use the AI Asset Registry within Unity Gateway, which is built on Unity Catalog. This unified registry gives organizations a central place to register, discover, secure, and audit all AI models, agents, MCPs (Model Context Protocol servers), tools, and skills using the same governance framework you use for data.
Instead of teams independently creating integrations and managing separate inventories, your platform team gets a single, authoritative inventory and developers get a catalog of trusted, discoverable building blocks. The registry becomes the single source of truth for your AI estate, preventing the "you can't govern an AI estate if you don't know what exists" problem.

What is the AI Asset Registry?

The AI Asset Registry is the foundation for both discovery and runtime enforcement of AI assets in Unity Gateway. It operates as an extension of Unity Catalog, meaning all AI assets live in Unity Catalog schemas and inherit standard metadata management. Each asset carries its name, owner, comment, and tags, making it easy to organize and track ownership across your organization.

Key capabilities for managing your AI inventory

Capability	Why it matters
Unified governance framework	Govern models, agents, MCP services, and skills using the same GRANT and REVOKE statements you use for tables and functions. No fragmented access systems.
Central discovery	All AI assets are discoverable in Catalog Explorer alongside the rest of your Unity Catalog assets, creating a single discovery experience.
Consistent access policies	Apply role-based access control consistently across all AI assets without maintaining separate access models.
Permissions and lineage	Track who has access to which assets, understand data dependencies, and audit which AI models and agents use which data sources.
Observability and cost tracking	Monitor how assets are used across teams and track costs per model, agent, or team to enforce budget controls.
Runtime enforcement	Unity Gateway enforces your governance policies not just at access time but during execution of model calls, tool invocations, and agent workflows.

How Databricks approaches AI inventory management

Databricks Unity Gateway provides several mechanisms to build and maintain your AI asset inventory:
Publish and reuse assets. Teams publish an AI asset once to the registry, assign appropriate permissions, and make it discoverable. Approved assets become reusable building blocks that users and agents can discover and call directly.
Integration with Unity Catalog. Because AI assets live in Unity Catalog schemas, you eliminate the need to maintain separate inventories or access models for AI. Your existing data governance skills and policies automatically extend to AI assets. Standard Unity Catalog metadata, owner information, and commenting ensure consistent documentation.
Centralized visibility. Administrators gain a complete view of the AI estate through permissions management, lineage tracking, and observability dashboards. You can see which models are used, by whom, and for what cost.
Governance at scale. As your use of AI models and agents grows, Unity Gateway applies consistent governance policies across all assets, preventing ad hoc integrations and shadow AI systems.

Use cases

Multi-team organizations. In large companies with dozens of data science and platform teams, the AI Asset Registry prevents duplicate model deployments and creates reusable assets that multiple teams can use. Platform teams use it to curate approved models and MCP integrations for the broader organization.
Regulated industries. Financial services, healthcare, and other regulated sectors use the inventory to demonstrate compliance through audit trails, access logs, and cost tracking for each model and agent.
Cost governance. Organizations with significant AI spending use the inventory to allocate costs by team, product, or business unit, then enforce budget caps on who can call which models.

FAQs

What types of assets can I inventory with the AI Asset Registry?

The registry supports all AI assets: foundation models, fine-tuned models, agents, MCP servers, tools, skills, and model services. This means your entire AI estate can be tracked in one place.

Can I apply different access policies to different models and agents?

Yes. Use standard Unity Catalog GRANT and REVOKE statements to assign permissions at a granular level. You can restrict access to sensitive models, allow specific teams to call certain agents, or make approved assets public within your organization.

How does Unity Gateway enforce governance policies at runtime?

Unity Gateway intercepts model calls, tool invocations, and agent workflows to check permissions, enforce rate limits, track usage, and ensure compliance with your access policies before allowing the action to complete.

Can I track which models are actually being used by my organization?

Yes. Unity Gateway provides observability and usage tracking so you can see which models are called by which applications or users, understand cost per asset, and identify unused models that can be deprecated.

Is there a Catalog Explorer interface to browse my AI assets?

Yes. All AI assets appear in Catalog Explorer alongside your data assets, providing a unified discovery experience. You can browse, search, and filter AI assets the same way you do tables and views.