Skip to main content

Knowledge Graph API Reference

All endpoints live under /api/v1/rdf unless noted. Every one returns 503 when RDF is disabled. Authenticate with a bearer token:
Most knowledge-graph endpoints are admin-only. Graph node hydration resolves entity details without re-applying the caller’s per-entity view authorization, so opening these to non-admins would leak entities a user cannot otherwise see. The glossary-scoped ontology endpoints on /api/v1/glossaries are the non-admin path.

Status

defaultLevel is NONE, and availableLevels lists NONE, plus CUSTOM when materialized inference is available. The status response uses uppercase, and the inference query parameter uses lowercase.

SPARQL

POST /sparql/update writes directly to a derived index. Anything you write is discarded on the next reindex. Use the entity APIs to change your catalog.

Entities and Graph

GET /glossary/graph Parameters

Passing both glossaryId and glossaryTermId requires the term to belong to that glossary while still returning direct cross-glossary neighbors.

Validation

Inference Rules

Custom Ontology Extensions

Custom classes and properties must be in https://open-metadata.org/ontology-extension/.

Insights

All admin-only.

SQL → SPARQL

Linked Open Data

Ontology Endpoints

Ontologies have their own endpoints under /api/v1/ontologies, /api/v1/ontologyReleases, and related paths. They work with RDF storage disabled, because they read the database-primary model. For more information, see Ontology API Reference.

Applications

Only one reindex job can be active per cluster at a time.

Next

MCP tools

The same capabilities, for agents.