Linked Data MCP

MCP server for AI agents to progressively access W3C Semantic Web specifications. Instead of dumping entire spec docs into the rotting context, it serves them section-by-section. Your AI can now look up that one SHACL 1.2 property it keeps forgetting about without burning 100k tokens.

Supported Standards

Family	Specifications	Namespace
RDF	RDF 1.1/1.2 Primer, Concepts, Semantics, Turtle, TriG, N-Triples, N-Quads, JSON-LD	rdf:, rdfs:
SPARQL	SPARQL 1.1/1.2 Query, Update, Overview	-
OWL	OWL 2 Primer, Overview, Syntax, Profiles, RDF Mapping	owl:
SHACL	SHACL 1.1/1.2 Core, Advanced Features, SPARQL, Node Expressions, Rules, UI	sh:
SKOS	SKOS Primer, Reference	skos:
PROV	PROV Primer, Data Model, Ontology	prov:
TIME	OWL-Time	time:

See the index.yaml file for the full list of specifications and namespaces. Please feel free to submit a PR to add more specifications and parsers.

Quick Start

Option 1: FastMCP Cloud (Recommended)

Fork the repo, configure the index.yaml and deploy to your own fastmcp.cloud instance (free for personal use). Afterwards, you can use their web ui to connect your agentic coding tool (e.g. Cursor, Claude Desktop, etc.) to the MCP server.

Option 2: Install as a tool

Prerequisites: Python 3.11+ and uv.

git clone https://github.com/robsyc/ld-mcp && cd ld-mcp
uv tool install .

Then add to your .cursor/mcp.json or Claude Desktop config:

{
  "mcpServers": {
    "ld-mcp": {
      "command": "ld-mcp"
    }
  }
}

Option 3: Run with uv

git clone https://github.com/robsyc/ld-mcp && cd ld-mcp
uv sync

Add to your .cursor/mcp.json or Claude Desktop config:

{
  "mcpServers": {
    "ld-mcp": {
      "command": "uv",
      "args": ["tool", "run", "ld-mcp"]
    }
  }
}

Available Tools

Tool	Description
list_specifications(family?)	Browse spec families (RDF, SPARQL, OWL, SHACL, SKOS, etc.)
list_sections(spec_key, depth?)	Get TOC with section IDs
get_section(spec_key, section_id)	Get markdown content for a section
list_resources(ns_key)	List classes/properties in a namespace
get_resource(ns_key, resource)	Get Turtle definition of a resource

Configuration

Variable	Description	Default
SPEC_VERSIONS	Filter by version (e.g., "1.2")	All
CACHE_TTL	Cache TTL in seconds	86400 (24 hours)
INDEX_PATH	Path to a custom index.yaml	Bundled default

Development

uv sync --group dev

# Validate all specs/namespaces
uv run pytest tests/test_index.py -v

# Lint
uv run ruff check src/ tests/

# MCP Inspector (web UI for testing tools)
uv run fastmcp dev inspector src/ld_mcp/server.py:mcp

Adding Specifications

1. Add entry to src/ld_mcp/index.yaml
2. Run uv run pytest tests/test_index.py -v -k "your_spec_key"
3. Push to main — CI validates automatically

Acknowledgments

• W3C for the specifications.
• html-to-markdown for converting HTML to Markdown.
• RDFLib for the RDF library.
• FastMCP for the MCP server.