fabio-rovai/open-ontologies

What it solves

Open Ontologies provides a comprehensive environment for AI-native ontology engineering. It addresses the difficulty of building, validating, and governing complex RDF/OWL ontologies by providing a set of specialized tools that allow LLMs (like Claude) to act as ontology engineers, replacing traditional, manual JVM-based tools like Protégé.

How it works

The project consists of a a Rust-based MCP (Model Context Protocol) server and a visual desktop Studio. The server exposes over 70 tools for operations such as validation, querying, reasoning, and alignment using an in-memory Oxigraph triple store. It employs a three-layer architecture (Dynamics $\rightarrow$ Causal $\rightarrow$ Planner) to handle action registration, causal certification, and PDDL-based planning. The Studio provides a visual interface with a virtualized tree for exploring large ontologies, an AI chat panel for natural language engineering, and a property inspector for manual triple editing.

Who it’s for

It is designed for knowledge engineers, AI researchers, and developers building knowledge graphs who want to leverage LLMs to automate the creation and governance of structured semantic data.

Highlights

• MCP-Native Architecture: Designed specifically for LLMs to call tools for building and validating ontologies without requiring internal LLM clients or API keys.
• Three-Layer Stack: Includes Dynamics for atomic ticks and causal laws, Causal for action certification, and Planner for PDDL compilation and validation.
• Virtualized Tree: A high-performance UI capable of rendering 1,500+ classes without lag.
• Extensive Toolset: 70+ tools covering everything from SHACL validation and OWL-EL classification to fuzzy-logic alignment and RAG evaluation.
• Lineage Audit Trail: A full history of every plan, apply, and align event stored in SQLite for complete transparency of AI actions.