Note AI Governance and AI Command Center leverage the same operating model. Therefore, the content in this topic is identical to the operating model content presented in the AI Governance topic.
In the Collibra 2026.06 release, we made the following updates to the operating model.
Important To align with the updated operating model, you may need to perform several post-upgrade configuration tasks. For a comprehensive checklist of steps needed to finalize your environment transition, go to the Required actions section.
New asset types
The following new out-of-the-box asset types are added, to deepen our AI platform integrations for AI Agents:
| New asset type | Description |
|---|---|
| AI Agent Version | A specific implementation of an AI agent at a point in time, capturing details like the tools, models, and configurations used. |
| AI Agent Tool | A service accessible to an AI agent that automates tasks, generates content, or delivers intelligent insights. |
AI Agent Version details
- The AI trust score is included on AI Agent Version asset pages.
- AI Agent Version assets are shown:
- In the Registry.
- In the Trust Score widget, for agents.
- In the Assets by lifecycle status widget, for agents.
Asset status: Missing from source
We added the "Missing from source" asset status to the retirement phase of the following asset types, because this status can be set via our supported AI integrations:
- AI Agent
- AI Agent Version
- AI Agent Tool
- AI Model Version
- AI Model Deployment
- AI Base Model
New complex relation type
A new complex relation type, AI Agent Deployment, is introduced. It acts as a 'light asset type', linking an AI Agent Version to the AI Endpoint where it is exposed, without requiring a distinct asset name or lifecycle status. This design choice was made because most platforms treat deployments as technical objects identified by their relations rather than by a standalone name. For more information on the rationale, go to Preparing for deeper AI Agent integrations.
Updates to existing asset types
For the AI Agent asset type, the Allow identical asset names per domain option is now enabled by default. This supports ingestion scenarios where multiple agents in the same domain share the same display name.
New relation types
Explicit relation types
The following explicit relation types are added to support the new AI Agent asset hierarchy:
| Description | Relation format | Public ID |
|---|---|---|
| Links the version-agnostic AI Agent concept to the concrete agent versions that were built. | AI Agent has version / is version of AI Agent Version | AgentHasVersion |
| Links an AI Agent Version to its deployment. | AI Agent Deployment is deployment of / is deployed by AI Agent Version | DeploymentOfAIAgentVersion |
| Links an AI Agent Deployment to the AI Endpoint where it is exposed. | AI Agent Deployment is exposed on / exposes AI Endpoint | DeploymentExposedOnAIEndpoint |
| Links an AI Agent Deployment to the AI Monitor(s) tracking its performance. | AI Agent Deployment is monitored by / monitors AI Monitor | DeploymentMonitoredByAIMonitor |
| Identifies the model endpoint that an AI Agent Version invokes at runtime for inference. Captures the operational dependency between a specific agent version and the model endpoint it uses to process requests. | AI Agent Version infers using LLM / is used for inference of AI Endpoint | AgentVersionInvokesModelEndpoint |
| Identifies the tool that an AI Agent Version can invoke at runtime. | AI Agent Version can call / can be called by AI Agent Tool | AgentVersionCallsTool |
| Identifies the subagent that an AI Agent Tool delegates requests to. Because the subagent may be linked via a specific agent version or via an endpoint depending on the platform, two relations are provided to capture both options. |
AI Agent Tool delegates to / can receive requests from AI Agent Version AI Agent Tool delegates to / can receive requests from AI Endpoint |
AIAgentToolCallsAIAgentVersion AIAgentToolCallsAIAgentEndpoint |
| Links an AI Agent Tool to the data source it can read. | AI Agent Tool can consult data source / can be consulted via Inference Data | AIAgentToolCanReadInferenceData |
| Links an AI Agent Tool to the data source where it can create, update, or delete information. | AI Agent Tool can manipulate data in / can be manipulated via Output Data | AIAgentToolCanManipulateOutputData |
Derived relation types
To bring as much relevant information as possible to the AI Agent asset page, and to reduce the need for users to navigate asset by asset, the following derived relation types are introduced. These relations are computed automatically by traversing the knowledge graph.
| Description | Relation format | Public ID |
|---|---|---|
| Indicates the AI Endpoints where the AI Agent can be accessed. Derived automatically by tracing the agent versions, identifying their deployments, and checking the AI Endpoints of those deployments. | AI Agent can be accessed through / provides access to AI Endpoint | AIAgentAccessibleThroughAIEndpoint |
| Indicates the active AI Monitors for the AI Agent. Derived automatically by tracing the agent versions, identifying their deployments, and checking the active AI Monitors on those deployments. | AI Agent is monitored by / monitors AI Monitor | AIAgentMonitoredByAIMonitor |
| Indicates the AI model powering the agent. Derived automatically by tracing the agent's versions, following the model endpoints used, to the model versions deployed at those endpoints, and ultimately to the AI Base Model driving the agent's behavior. | AI Agent infers using LLM / is used for inference of AI Model | AgentInvokesModel |
| Links a calling agent to the subagent it can use. | AI Agent uses / is used by AI Agent | AIAgentCallsAIAgent |
| Links an AI Agent to the data sources it can read. | AI Agent can consult data source / can be consulted via Inference Data | AIAgentCanReadDataSource |
| Links an AI Agent to the data sources where it can create, update, or delete information. | AI Agent can manipulate data in / can be manipulated via Output Data | AIAgentCanManipulateDataSource |
New attribute types
| Attribute type | Description | Assigned to asset type |
|---|---|---|
| Traffic Split Percentage | The percentage of traffic on the endpoint that is being diverted to this deployment. | AI Agent Deployment |
| Endpoint Availability | Controls how the model endpoint is deployed and billed. Allowed values: Online, Serverless, Provisioned Throughput, Batch Processing, Asynchronous, Developer / Experimental. | AI Endpoint |
| Supported Interaction Protocols | Lists the interaction protocols supported by this endpoint. Allowed values: REST/HTTP, gRPC, MCP, A2A, ACP, ANP, LMOS, AG-UI, A2UI. | AI Endpoint |
Preparing for deeper AI Agent integrations
These operating model changes prepare AI Governance for richer integrations that bring in more information about AI Agents: the models and system instructions they use, the tool calls they make, and the systems they interact with.
AI Agent Deployment as a complex relation type
Learning from challenges encountered when building integrations with the AI Model Deployment asset type, AI Agent Deployment is introduced as a complex relation type rather than a full asset type. This 'light asset type' approach requires no asset name or lifecycle status, which better reflects how most platforms treat deployments: as technical objects identified by their relations to an agent version and an endpoint, often without a distinct name. Given the limited set of attributes and relations needed, a complex relation type is the more appropriate fit.
The rise of derived relation types
While asset types such as AI Agent Version, AI Endpoint, and the AI Agent Deployment complex relation type are needed to store information accurately, navigating this information page by page can be cumbersome for end users. The new derived relation types address this by surfacing richer context directly on the AI Agent asset page, allowing the system to traverse the knowledge graph on the user's behalf.
Expected future evolutions
No attribute types or relation types are being deprecated in this release. The following changes are planned for future releases as we roll out updated integrations. Details may change as integrations evolve; refer to release notes for upcoming changes.
- The Instructions attribute type on AI Agent will move to AI Agent Version, so that the evolution of the system prompt can be captured over time.
- The Tool Usage attribute type on AI Agent will be replaced by relations to AI Agent Tool assets, which can then be linked to data sources, subagents, or other systems.
- The relation type AI Agent uses / is used by AI Model Version will be replaced by linking the model endpoint via AI Agent Version infers using LLM / is used for inference of AI Endpoint. A derived relation type, AI Agent infers using LLM / is used for inference of AI Model, will allow consumer users to understand which model an agent uses without needing to navigate to the endpoint level.
- Future extensions for MCP tools and skills can also be expected.
Recommended actions
Ensure that you read through the following points and take action as necessary.
Re-run your integrations
Monitor the release notes for your integrations. When an integration expands to ingest more data, review its configuration to determine whether new options require enabling or disabling. Then re-run the integration on demand or wait for the next scheduled sync.
Evaluate your custom workflows
For new asset types, relation types, and attribute types, no specific action is required. If you have custom workflows or integrations, evaluate whether the new types could enhance your logic.
Note To ensure workflow stability, we recommend using UUIDs or public IDs rather than display names when retrieving asset, relation, or attribute types. Unlike names, these identifiers remain constant even if aspects of the operating model are renamed.
Evaluate customized asset page layouts
We updated the default asset page layouts for AI Governance asset types to reflect all asset model changes. If you customized a layout, it is unsubscribed from the default and remains unchanged to preserve your customizations. We recommend reviewing customized layouts to align them with these asset model changes. You can choose one of two approaches:
- If you only reordered characteristics or changed sections, consider reverting to the default layout to resubscribe to future automatic updates.
- If you added custom characteristic types that you want to keep, manually add new out-of-the-box relation types and attribute types to the layout.