Standards Comparison
UKS doesn't replace existing standards — it carries and bridges them. An L1 packet interoperates with citation and dataset formats today; the registry emits several of them directly. Here's how UKS relates to the standards you already use.
At a glance
| Standard | Scope | Relationship to UKS |
|---|---|---|
| schema.org / JSON-LD | Web-wide structured data | UKS packets serve as Dataset JSON-LD via content negotiation on /packets/:id; entities map to schema.org types |
| MLCommons Croissant | ML dataset metadata | /packets/:id/croissant emits Croissant → discoverable in HF, Kaggle, Google Dataset Search |
| schema.org ClaimReview | Fact-check markup | /claims.jsonld emits a ClaimReview feed (contested-by-default) |
| FHIR | Health data exchange | Health-domain sources export to FHIR-shaped records via the SDK/CLI |
| BibTeX / RIS | Bibliographic citation | uks convert imports BibTeX/RIS into L1 sources |
| GRADE | Evidence quality | UKS adopts GRADE directly as the closed evidence_grade vocabulary |
| MCP | Agent ↔ tool protocol | @uks/mcp-server serves packets + the registry as MCP tools |
What UKS adds that others don't
Most formats describe data. UKS additionally carries, in one portable document:
- Evidence + clinical status as first-class, never-collapsed fields (GRADE-aligned) — schema.org/Croissant have no equivalent.
- License clarity per record (
license_label+rights_url) driving a fail-closed reuse decision. - Acquisition + extraction (Layers B/C) — how to get and parse more, not just what you have.
- Agent actions (Layer E) — what an agent should do with the knowledge, portably, outside any one orchestration framework.
Positioning
Think of UKS as the envelope that wraps your existing citation/dataset metadata and adds the trust, licensing, and agent-action layers that AI consumption needs — while still emitting the established formats (JSON-LD, Croissant, ClaimReview, FHIR) on demand for interoperability.
→ Related: What is UKS? · Query the registry · JSON Schema