Roadmap
This roadmap evolves with the project.
Phase 1: OntoCore
Status: Complete
The foundation. OntoCore is our neuro-symbolic core that transforms natural language skill definitions into validated OWL 2 DL ontologies.
- Natural language parsing with Claude
- OWL 2 DL serialization (RDF/Turtle)
- SHACL validation gatekeeper
- Security audit pipeline
- 156+ tests
Phase 2: OntoSkills
Status: Complete
The knowledge base. OntoSkills are the compiled, validated skills published from OntoCore — ready to be queried by agents.
- Core skill library compilation
- Public skill store
- Skill versioning and updates
- Dependency management
Phase 3: OntoMCP
Status: Complete
The interface. OntoMCP exposes OntoSkills via the Model Context Protocol, giving any MCP-compatible agent instant access to structured knowledge with sub-millisecond SPARQL queries.
- Rust MCP server with stdio transport
- Oxigraph in-memory graph store
- SPARQL 1.1 query interface
- Unified ontoskill tool (search + context retrieval + knowledge querying in one call)
- Per-skill pre-computed embeddings (optional, requires ontocore[embeddings])
- Embedding download at install time (—with-embeddings flag)
- Category and is_user_invocable search filters
- Claude Code integration
Phase 4: OntoStore
Status: In Development
The marketplace. OntoStore is a centralized repository where teams can publish, discover, and share ontologies.
- Per-package manifests with embedding file references
- Registry index with global embedding model declaration
- Ontology store with live marketplace
- 3D knowledge graph visualization
- Version management
- Team collaboration features
- Community contributions
Phase 5: OntoClaw
Status: Planned
The agent. An autonomous agent powered by structured knowledge — reasoning with precision, not hallucination.
- Agent architecture design
- Multi-agent collaboration
- Knowledge graph reasoning
- Production deployment
Track progress
Follow development on GitHub.
Have ideas? Open an issue.