Skip to content

Knowledge Graph for Education Global MCP

KGForEdGlobalMCP is a curriculum-agnostic, read-only FastMCP server for exploring versioned curriculum knowledge graphs from countries, states, and educational organizations.

It provides one MCP interface for discovering curriculum frameworks, searching and retrieving academic standards, navigating source hierarchies, accessing provenance-aware resources, and assembling bounded evidence for cross-framework comparison and progression analysis.

Deterministic server, client-side reasoning

The server loads, validates, indexes, and retrieves source-grounded curriculum evidence. It does not call a server-side LLM. Reasoning and generated outputs are performed by the connected MCP client or host model, such as Claude Desktop.

  • Search curriculum standards

    Discover frameworks and search package-local curriculum content using lexical text search or profile-governed code search.

    Search and retrieve standards

  • Navigate source hierarchies

    Retrieve direct relationships, bounded ancestors and descendants, and complete bounded paths to framework roots.

    Navigate hierarchies

  • Compare bounded evidence

    Retrieve independently scoped evidence across frameworks without treating retrieval results as official alignment or equivalence.

    Compare framework evidence

  • Use role-oriented workflows

    Use deterministic prompt workflows for student support, teacher materials, progression hypotheses, and administrator review.

    Use prompt workflows

What the server exposes

The current MCP surface contains the following:

Surface Count Purpose
Tools 13 Framework discovery, standards and learning-component retrieval, graph context, statistics, capabilities, comparison, progression
Fixed resources 1 Server catalog
Resource templates 12 Framework, package, validation, profile, standards, learning components, provenance, relationship, unresolved, and artifacts
Prompts 7 Role-oriented and cross-framework client-side workflows

Tools

  • list_frameworks
  • get_framework
  • search_standards
  • get_standard
  • get_standard_context
  • search_learning_components
  • get_learning_component
  • get_learning_component_context
  • get_learning_components_for_standard
  • get_framework_statistics
  • get_capabilities
  • compare_framework_evidence
  • collect_progression_evidence

Prompts

  • student_study_support
  • teacher_guide_draft
  • student_handbook_section
  • multigrade_lesson_plan
  • inferred_progression_hypothesis
  • administrator_alignment_review
  • cross_framework_comparison

See the MCP reference for the complete public surface.

Current curriculum catalog

The repository snapshot currently contains six accepted Academic Standards graph packages:

Framework Jurisdiction Subject Local grades or stages Code search
English Language Curriculum for Primary Schools (Basic 1 - 3) Ghana English Language BASIC 1, BASIC 2, BASIC 3 Partial
Mathematics Curriculum for Primary Schools (Basic 4 - 6) Ghana Mathematics BASIC 4, BASIC 5, BASIC 6 Partial
Learning Framework - Science India Science Class IX, Class X Partial
Tamil Nadu Curriculum Framework 2025: Mathematics - Classes 1-5 Tamil Nadu, India Mathematics Class-1 through Class-5 Partial
9-Year Basic Education Mathematics Curriculum for Primary 1-3 Nigeria Mathematics PRIMARY ONE through PRIMARY THREE None
Mathematics Syllabus for Lower Primary | Primary 1 - 3 Rwanda Mathematics P1, P2, P3 None

All six packages support lexical text search and include detailed provenance. Search and traversal capabilities remain package-specific; use get_capabilities to inspect the runtime for supported modes for each framework.

For package-level details, see Available frameworks.

How it works

Each curriculum is retained as an independent, immutable graph package. Shared services make those packages discoverable through one server without merging their source structures into a single curriculum graph.

%%{init: {"themeVariables": {"fontSize": "18px"}}}%%
flowchart LR
    A[Interpretation profiles] --> D[Validated application runtime]
    B[Prompt configurations] --> D
    C[Versioned graph packages] --> D
    D --> E[Package-local graph stores]
    D --> F[Package-local search indexes]
    D --> G[Rights-aware resources]
    D --> H[Deterministic prompt service]
    E --> I[FastMCP tools and resources]
    F --> I
    G --> I
    H --> J[FastMCP prompts]
    I --> K[MCP client / host model]
    J --> K
Hold "Alt" / "Option" to enable pan & zoom

This architecture preserves:

  • exact framework, snapshot, and graph-package identity;
  • source terminology and local grade labels;
  • provenance, rights, and validation evidence;
  • package-local tree or multi-parent DAG topology; and
  • unresolved relationship statuses where the source package retains them.

Check out Architecture for the runtime design and Concepts and boundaries for the interpretation rules behind these guarantees.

Important boundaries

KGForEdGlobalMCP is designed to return faithful, bounded evidence rather than silently infer educational claims.

Graph hierarchy is not learning progression

A structural hasChild relationship records retained source hierarchy. It does not, by itself, assert prerequisite knowledge, instructional sequence, conceptual dependency, or learner mastery.

Framework comparison is not curriculum alignment

Cross-framework retrieval returns candidate evidence from independently bounded searches. It does not establish official alignment, equivalence, grade equivalence, or endorsement between frameworks.

Text search is lexical

Text search uses normalized tokens or contiguous normalized phrases. It does not perform semantic search, stemming, lemmatization, or automatic synonym expansion. A zero-result query means the supplied lexical expression did not match under the requested bounds; it does not necessarily mean that the curriculum lacks the concept.

The current server also does not provide persisted alignments, accepted mapping overlays, embeddings, semantic retrieval, snapshot diffs, server-side LLM calls, or first-class learning-progression graphs.

Where to start

Next: Architecture