Skip to main content

Data Catalog

The Data Catalog provides a unified view of all AI entities discovered from your integrations, including agents, models, tools, downstream systems, and identities.

Overview​

Access the Data Catalog from the main navigation sidebar. It displays all discovered entities with their type, source, approval status, and tags.

Grid View​

Data Catalog Grid ViewData Catalog Grid View

The grid view displays entities as cards showing the entity name, path, department, deployment stage, type badge, and approval status. Use the filters panel on the left to narrow results by resource type or tags.

List View​

Data Catalog List ViewData Catalog List View

Switch to list view for a sortable table format showing entity name, type, source, and tags columns.

Graph View​

Data Catalog Graph ViewData Catalog Graph View

The graph view visualizes relationships between entities — agents, tools, models, and downstream systems.

Graph Controls​

  • Zoom — Scroll to zoom in/out, or use the +/- controls
  • Pan — Click and drag the background to pan the view
  • Select — Click a node to select it and view its details
  • Recenter — Click the recenter button to reset the view to fit all nodes

Legend​

The graph legend (displayed in the corner of the graph view) explains the visual encoding:

  • Node types — Different shapes or colors represent agents, models, tools, downstream systems, and identities
  • Edge types — Lines represent relationships (e.g., "Uses")
  • Approval status — Color coding indicates status: Approved (green), Under Review (yellow), or Not Approved (red)

Entity Detail Panel​

Click any node in the graph to open its detail panel, which shows:

  • Entity name and type
  • Approval status and deployment stage
  • Connected entities — What it connects to in the graph
  • Tags applied to the entity
  • Recent activity — Last seen and trace count

Model Cost Information​

Click a model node to see cost details:

  • Cost per 1K input tokens and output tokens (from Cost Estimates)
  • Total estimated cost for the selected time period
  • Token usage — Total input and output tokens consumed

Use the search bar to find entities by name:

🔍 Search entities...

Resource Type Filters​

Filter entities by type using the checkboxes in the Filters panel:

  • Model — LLM models (e.g., gpt-4o-mini, claude-3-opus)
  • Tool — Tools and functions called by agents
  • Agent — AI agents discovered from traces
  • MCP Server — Model Context Protocol servers
  • Downstream System — External systems agents interact with
  • Identity — Non-human identities associated with agents

Tag Filters​

Filter by tags applied to entities. Tags can be managed from the Tags settings page.

Entity Properties​

Each entity in the catalog has the following properties:

PropertyDescription
NameThe entity's display name
PathThe fully qualified path (e.g., discovery/agents/CustomerSupport)
TypeAgent, Model, Tool, MCP Server, Downstream System, or Identity
SourceHow the entity was discovered (e.g., Auto Discovery)
DepartmentOrganizational department assignment
StageDeployment stage (e.g., Unknown, Dev, Staging, Prod)
Approval StatusApproved, Under Review, or Not Approved
TagsLabels for categorization

Approval Status​

Each entity has an approval status that indicates whether it has been reviewed and sanctioned for use:

StatusMeaning
ApprovedReviewed and approved for use
Under ReviewPending approval from governance team
Not ApprovedNot yet reviewed or explicitly blocked

Approval status is used by policies like Unapproved Tool Use and Approved Model Version to enforce governance rules.

Show Details​

Use the Show Details toggle in the top-right corner to expand or collapse additional metadata on entity cards in the grid view, such as department, stage, and description fields.

Integration with Policies​

The Data Catalog works closely with governance policies:

  • Unapproved Tool Use — Flags tools that are not marked as Approved in the catalog
  • Approved Model Version — Ensures only approved models are used in production
  • Cross-Boundary Access — Uses the Stage property to enforce environment isolation

Best Practices​

1. Review Discovered Entities​

Regularly review newly discovered entities and set their approval status appropriately.

2. Use Consistent Tags​

Establish tagging conventions:

  • Environment: production, staging, development
  • Team: team:analytics, team:ml
  • Compliance: pii, confidential

3. Set Deployment Stages​

Assign accurate deployment stages to entities so cross-boundary access policies can enforce environment isolation.

4. Approve Sanctioned Assets​

Mark reviewed entities as Approved so governance policies can distinguish sanctioned from unsanctioned usage.

Next Steps​