feat: add DataHub MCP server

[Timur99]

What

Adds a catalog entry for the official DataHub MCP server (datahub.yaml).

DataHub is a metadata platform for the modern data stack. The server lets an agent search the data catalog, trace table- and column-level lineage, inspect dataset schemas, look at the SQL that actually runs against a table, and curate metadata such as tags, glossary terms, owners, domains, and structured properties.

Why remote rather than containerized

The upstream project is primarily distributed as a local stdio server (uvx mcp-server-datahub), which this catalog does not support. Rather than requiring a new image in obot-platform/mcp-images, this entry targets DataHub's managed MCP endpoint, which is served by every DataHub deployment at /integrations/ai/mcp:

• DataHub Cloud: https://<tenant>.acryl.io/integrations/ai/mcp
• Self-hosted DataHub: the same path on the deployment's own domain

Because the host differs per user, the entry uses the URLTemplate + bearer-header pattern already established by databricks_genie.yaml, databricks_uc_functions.yaml, databricks_vector_search.yaml, and postman.yaml:

runtime: remote
remoteConfig:
  URLTemplate: ${DATAHUB_URL}/integrations/ai/mcp
  headers:
  - name: Personal Access Token
    description: DataHub personal access token
    key: Authorization
    required: true
    sensitive: true
    prefix: "Bearer "

This keeps the entry free of any image-build dependency and works for Cloud and self-hosted users alike.

Configuration

Input	Type	Required	Notes
DATAHUB_URL	env	yes	Base URL of the DataHub deployment, no trailing slash
Authorization	header	yes	DataHub personal access token, sent as Bearer <token>

Docs reference: DataHub MCP Server. The managed endpoint requires DataHub Cloud v0.3.12+; metadata curation tools require v0.3.17+ and are enabled server-side, so tools that are disabled on a given deployment simply won't appear.

Tool preview

The 21 previewed tools were taken from the tool registrations in src/mcp_server_datahub/mcp_server.py and the corresponding function signatures in src/mcp_server_datahub/tools/*.py upstream, so names, parameters, and defaults match the current server surface rather than being paraphrased from the README.

Read tools: search, get_lineage, get_dataset_queries, get_entities, list_schema_fields, get_lineage_paths_between, search_documents, grep_documents, get_me.

Curation tools: add_tags, remove_tags, add_terms, remove_terms, add_owners, remove_owners, set_domains, remove_domains, update_description, add_structured_properties, remove_structured_properties, save_document.

Notes

• Single new file; no existing entries touched.
• entryKey: obot-datahub is unique across the catalog.
• Categories reuse the existing vocabulary: Data & Analytics, Databases, Developer Tools.
• Happy to also add datahub.yaml to the choice list in .github/workflows/tool-preview-refresh.yml if you'd like — note that list is currently missing several existing entries, so I left it alone to keep the diff minimal.