Universal Ontology MCP v0.1

The Intelligent Bridge between Unstructured Data and High-Fidelity Knowledge Graphs.

Universal Ontology MCP is a powerful tool designed for AI assistants to explore, navigate, and populate complex ontologies. It transforms raw text into structured relationships while adhering to strict semantic standards.

🚀 Why Universal Ontology MCP?

Existing ontology tools often struggle with semantic ambiguity and rigid keyword matching. This MCP solves these problems by providing:

🧠 Balanced Semantic Search: Don't get stuck on exact names. We use state-of-the-art embeddings (default: all-MiniLM-L6-v2) with a 50/50 weighted scoring between Class Names and Descriptions for maximum intuitive accuracy.
⚡ Proactive Schema Guidance: The server doesn't just list properties; it teaches the AI how to use them. It identifies mandatory fields and expected entity types for ObjectProperties in real-time.
🏗 Component-Based Modeling: Simplifies complex modeling (like UCO Facets) by ranking and recommending relevant components for any given class.
🔍 ASCII Graph Visualization: Preview your knowledge graph structure directly in the chat interface before exporting, with support for both clean and detailed (full URI) modes.
⚖️ Built-in SHACL Validation: Ensures data integrity from the start. It validates entities against schema constraints before you export your graph.
� Session Control & Sanitization: surgically remove entities, reset the session environment, and enjoy automatic URI sanitization for complex IDs (like file paths or registry keys).

🌟 Intelligent Tools

get_ontology_summary: Quick high-level overview of the loaded schema.
search_classes / search_properties: Weighted semantic discovery.
get_class_details: Detailed usage instructions & connectivity rules.
list_available_facets: Smart ranking of components for complex data grouping.
create_entity / set_property / attach_component: Atomic graph construction with auto URI sanitization.
remove_entity / reset_graph: Surgical modification or full session cleanup.
validate_entity: Instant SHACL compliance check.
visualize_graph: ASCII tree preview of the current graph (supports verbose mode).
export_graph: Save your validated knowledge graph to .ttl.

🏗 Architecture

main.py: Entry point for the MCP server.
mcp_server/engine.py: Core logic for ontology parsing, caching, and vector embedding calculations.
mcp_server/server.py: Tool definitions and FastMCP server configuration.
mcp_server/config.py: Persona instructions and environment defaults.

🛠 Installation

Clone the repository.
Install dependencies:
```
pip install -r requirements.txt
```
Set your ontology directory (path containing your .ttl files):
```
export ONTOLOGY_DIR="/path/to/your/ontology/folder"
```

🔌 MCP Configuration

Add this configuration to your MCP-compatible client (e.g., Gemini, Claude Desktop, VS Code).

Configuration Template

gemini-cli, claude(Desktop,Code)

{
  "mcpServers": {
    "universal-ontology-mcp": {
      "command": "python",
      "args": ["/absolute/path/to/universal-ontology-mcp/main.py"],
      "env": {
        "ONTOLOGY_DIR": "/absolute/path/to/your/ontology/folder",
        "EMBEDDING_MODEL": "all-MiniLM-L6-v2"
      }
    }
  }
}

OpenCode

{
  "mcp": {
    "universal-ontology-mcp": {
      "type": "local",
      "command": ["python","/Users/seonwookim/Documents/Programming/Security/cyber-ontology/ontology-mapper/universal-ontology-mcp/main.py"],
      "environment": {
        "ONTOLOGY_DIR": "/Users/seonwookim/Documents/Programming/Security/cyber-ontology/ontology-mapper/ontology",
        "EMBEDDING_MODEL": "all-MiniLM-L6-v2"
      }
    }
  }
}

⚖️ License

This project is licensed under the MIT License.

Universal Ontology MCP