Open knowledge infrastructure / Developer portal

Build with the language
of digital learning.

Query curated concepts, navigate semantic relationships, and ground AI applications in source-attributed knowledge. One public glossary, five interoperable interfaces.

Public & read-onlyCanonical glossary v0.0.88Persistent HTTPS identifiers

Five ways into the graph.

Start with your application, not an export format.

AI / SEMANTIC GROUNDING

Model Context Protocol

Connect AI assistants to the published glossary through four bounded, citation-aware, read-only tools. Public remote Streamable HTTP; no credentials required.

Set up MCP →
02 / JSON

REST API

Search concepts, retrieve definitions, traverse relationships, and request compact AI context bundles.

Explore REST →
03 / RDF

Linked Data

Use canonical concept URIs and machine-readable JSON-LD or Turtle representations.

Explore Linked Data →
04 / QUERY

SPARQL

Query the published RDF graph using familiar semantic-web query patterns.

Explore SPARQL →
05 / VISUAL

Graph Explorer

Inspect concepts, sources, and relationships visually before building a query or integration.

Open Explorer →

Give an assistant a reliable reference.

MCP clients supporting remote Streamable HTTP can discover and invoke the glossary's tools. The endpoint is POST-only; opening it in a browser returns HTTP 405 by design.

Connection

Add a remote MCP server in your AI client's settings, select Streamable HTTP, and use no authentication.

https://id.i2idl.org/mcp
  1. Register the URL as a remote Streamable HTTP MCP server.
  2. Discover the four read-only tools.
  3. Try get_context for learning-engineering.
  4. Preserve citations and canonical identifiers when generating an answer.

Tools available

Each tool retrieves only published reference material and declares itself read-only.

  • search_conceptsFind published terms and canonical IDs.
  • get_conceptRetrieve a concept, definition, and metadata.
  • get_relationshipsRetrieve asserted broader, narrower, and related concepts.
  • get_contextRetrieve a bounded semantic context bundle with provenance.
Verified with ChatGPT and MCP Inspector. Other client support depends on transport and protocol compatibility.
EXAMPLE / TOOL ARGUMENTS
{"name":"get_context","arguments":{"id":"learning-engineering","depth":1,"maxRelated":3}}

MCP Registry: io.github.blakeplock/i2idl-glossary · Registry version 1.0.0 · Negotiated MCP protocol 2025-06-18

Straightforward JSON for applications.

Search concepts

Use the public versioned API to retrieve a bounded search result.

HTTP GET
GET /api/v1/concepts?q=learning&limit=10

Retrieve grounded context

Get a single concept's definition, evidence, neighbors, and provenance.

HTTP GET
GET /api/v1/concepts/learning-engineering/context?depth=1&maxRelated=3

Durable identifiers. Open formats.

The canonical graph uses JSON-LD and RDF vocabularies including SKOS, Dublin Core Terms, Schema.org, and PROV-O. Concepts resolve via stable HTTPS identifiers.

ResourceExampleUse
Canonical identifier/concepts/learning-engineeringPersistent concept URI
JSON-LD/concepts/learning-engineering.jsonldMachine-readable concept graph
Turtle/concepts/learning-engineering.ttlRDF serialisation
Full graph/glossary.jsonldComplete publication

Ask questions of the RDF graph.

Query endpoint

The hosted endpoint supports bounded, read-only SELECT, ASK, CONSTRUCT, and DESCRIBE queries.

https://id.i2idl.org/sparql

Example query

SPARQL / SELECT
PREFIX skos: <http://www.w3.org/2004/02/skos/core#>
SELECT ?concept ?label WHERE {
  ?concept a skos:Concept ;
           skos:prefLabel ?label .
} LIMIT 10

Read operations only. Query execution is bounded by the service.

See the connections before you query.

The Knowledge Graph Explorer lets you inspect semantic relationships, compare concepts, study source coverage, and discover scholarship linked to glossary topics. Research discovery results are separate from editorially curated graph assertions.

Reference data with a clear provenance trail.

I2IDL concept records identify published definitions, relationships, citations, and source evidence. AI-generated explanations should distinguish explicit graph assertions from model interpretation. The public MCP service is unauthenticated and read-only; request bounds, upstream timeouts, and edge rate limiting protect service availability.