The Catalog Discovery MCP server allows your AI agents to discover and explore enterprise data assets in Data Governance and Catalog.
Use the Catalog Discovery MCP server to discover and explore enterprise data assets within Data Governance and Catalog. Your AI agents can search your catalog using natural language prompts, structured queries, or keywords, retrieve comprehensive asset metadata, and retrieve relationships between data assets. Use the Catalog Discovery MCP server to ground your LLM responses, ensuring that your AI works only with governed, accurate enterprise data context.
For more information about constructing structured queries, see Construction of search queries in the Data Governance and Catalog help.
Key capabilities
The Catalog Discovery MCP server offers the following key capabilities:
•Natural language search. Discovers data assets using conversational queries. The AI interprets intent, relationships, and implicit filters from plain English.
•Asset details. Retrieves full 360-degree metadata for any asset including stakeholders resolved to names or emails, classifications, glossary terms, hierarchy, and neighbourhood relationships.
•Keyword or query search. Discover data assets using structured queries or keywords with typed filters, aggregations for faceted exploration, and multi-field sorting. Examples of typed filters are classType, resourceType, certification, lifecycle, stakeholders, and dates.
Use cases
Use the Catalog Discovery MCP server to address the following use cases:
•Data discovery. Find relevant data assets across your catalog using natural language or structured queries.
•Data governance reporting. Quickly identify and report on assets by classification, certification status, or stakeholder ownership.
•LLM response grounding. Provide your AI with governed enterprise data context to prevent hallucination. Examples of context are names, types, and descriptions.
•Relationship exploration. Understand how assets connect through hierarchy, neighborhood relationships, and glossary linkages.
Tools
The following table describes the tools available on the Catalog Discovery MCP server:
search_assets
Searches for assets in your catalog using keyword search or natural language (NL) search.
Keyword mode is the default. Use it first.
Use keyword mode with filterSpec for precise structured queries.
By default, keyword searches for "contains" matches in common fields.
Keyword mode searches across the following fields:
- core.name
- com.infa.ccgf.models.governance.AliasNames
- core.businessName
- core.description
- core.technicalDescription
- core.externalIdentity
If keyword mode does not return expected results, retry with NL mode. You can also provide filters within a natural language (NL) query. For example, you can include the filter "show Oracle tables modified after 2025-01-20."
The following are NL query examples:
- "domain named finance"
- "Assets with classifications"
- "Assets with Glossaries"
- "Show me classifications"
- "Show me Glossaries"
- "Oracle Tables"
- "find the GDPR policy"
- "show BI reports related to policy GDPR"
- "show me business terms related to tables"
- "show columns related to business term Phone Number"
- "show datasets related to business term Customer"
- "show columns from table customers"
- "show data elements from dataset finance"
- "find data elements in the EMPLOYEE dataset"
- "find folders with customer files"
- "show everything in domain HIPAA"
- "show domains linked to policy GDPR"
- "show datasets not linked to business term"
- "show data elements not connected to data classification"
- "my business terms"
- "show domains certified by me"
- "show tables modified by me"
- "show business terms that contain columns and are related to data elements"
- "show all items named cluster or updated by system"
- "show business terms where lifecycle is In Review or Draft"
- "show assets modified after 2025-01-20"
- "find Payroll subdomains with Salary business terms"
Use * as a query to search your entire catalog and filter or aggregate results.
This query returns resolved asset summaries with the following attributes:
- identity
- class_type
- name
- description
- stakeholders
- asset_groups
The response includes aggregations for narrowing results and suggested_next_steps for iterative refinement. Use get_asset_details to dig deeper into a specific asset returned by this search.
Data Governance and Catalog organizes assets into the following categories:
- Glossary. Business vocabulary and definitions. The glossary includes Domain, subdomain, business term, and Metric.
- Data catalog: Technical assets retrieved from your catalog sources. Examples of technical assets are tables, views, columns, schemas, and files.
A regular hierarchy includes core.Resource > core.DataSource(Database/Schema) > core.Dataset(Table) > core.DataElement(Column).
Filter by core.ResourceType to filter assets belonging to a particular source system. Examples of resource types are Oracle and Amazon S3.
The following are high-level business asset types:
- System and data set. Business-friendly representations of technical infrastructure. Examples are System, data set, and data element.
- Process. Business activities and workflows.
- Policy. Internal rules and data governance guidance.
- Project. Planned activities to achieve governance objectives.
- AI. ML models and systems. Examples are AI model and AI system.
- Business Area, Legal Entity, Geography, Regulation
get_asset_details
Fetches full details of an asset by identity.
Use search_assets first to discover asset identities.
Returns a summary, which includes the following system attributes:
- identity
- name
- type
- location
- description
- stakeholders resolved to names or email addresses
- asset groups
- timestamps
Use segments to request the following additional aspects:
- summary
- selfAttributes
- stakeholdership
- dataClassification
- glossary
- hierarchy
- neighborhood
Note:
In AI Agent Engineering, the MCP server tools are called "actions."