A Model Context Protocol (MCP) server for DataHub that provides tools for searching, querying, and managing metadata entities in a DataHub instance
The DataHub MCP server has 23 tools with generally adequate descriptions and consistent naming patterns. Most tools start with action verbs (search, get, list, add, remove, update, set). However, there are significant gaps in parameter documentation, schema completeness, and output structure documentation. Many parameters lack explicit type definitions or constraints in the visible code. Error handling guidance is absent. The server demonstrates moderate quality suitable for internal use but lacks the rigor expected of production-grade agent tools.
Add owners to DataHub entities
Add structured properties to DataHub entities
Add tags to DataHub entities
Add glossary terms to DataHub entities
Enhanced search across DataHub entities with semantic search capabilities
Get data quality assertions for a dataset
Get queries associated with a dataset
Output schemas not documented. Tool descriptions state WHAT they do but do not specify the structure of returned data. LLMs cannot plan downstream tool calls or extract specific fields without knowing output structure.
No error handling guidance. Tools do not document what errors can occur, how to classify them (retryable vs user-fixable vs fatal), or what the agent should do on failure. Error responses lack actionable recovery instructions.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 64 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 35 | - | v1 |
Fetch detailed metadata for DataHub entities by their URNs
Get lineage information for a DataHub entity showing upstream and downstream dependencies
Get lineage paths between two DataHub entities
Get information about the authenticated user
Search document content using grep-like pattern matching
List all schema fields for a dataset
Remove domain assignment from multiple DataHub entities. This tool allows you to unset the domain for multiple entities in a single operation. Useful for removing domain assignments when reorganizing entities or correcting misassignments.
Remove owners from DataHub entities
Remove structured properties from DataHub entities
Remove tags from DataHub entities
Remove glossary terms from DataHub entities
Save a document to DataHub
Search across DataHub entities using full-text search with optional filters
Search for documents in DataHub catalog
Set domain for multiple DataHub entities. This tool allows you to assign a domain to multiple entities in a single operation. Useful for organizing datasets, dashboards, and other entities into logical business domains.
Update descriptions for DataHub entities
Parameter type descriptions are present in input schema but lack format constraints and validation rules in plain text. For example, 'num_results' is integer but no min/max bounds documented; 'filter' is string but no syntax guidance provided. LLMs cannot infer valid ranges or formats from bare type declarations.
Destructive write operations (add_tags, remove_tags, add_terms, remove_terms, update_description, set_domains) lack confirmation or dry-run capability. Agents may execute irreversible changes without validation, risking data corruption or metadata loss.
Pagination not clearly documented. search and enhanced_search accept num_results but do not describe offset/limit mechanics, total count, or whether results are paginated or capped. Large result sets may be truncated silently.
Batch operations accept lists (e.g., tag_urns, entity_urns) but do not document per-item success/failure reporting. If one entity fails to tag, the entire operation may fail, leaving agent in inconsistent state.
Parameter 'urns' in get_entities declared as 'string or list', ambiguous type. JSON Schema should declare union explicitly (oneOf: [string, array]). Ambiguous parameter types cause LLMs to guess representation, increasing failure rate.
No document about natural identifiers. Many tools accept entity_urns and dataset_urns (opaque system IDs). No indication whether human-friendly names, display identifiers, or lookups are supported. Forces users/agents to perform extra discovery calls.