name: open-ontologies version: "0.5.1" description: > AI-native ontology engineering using 110 MCP tools backed by an in-memory Oxigraph triple store. Build, validate, query, and govern RDF/OWL ontologies with a generate-validate-iterate loop. Use when building ontologies, knowledge graphs, RDF data, SPARQL queries, BORO/4D modeling, SHACL validation, clinical terminology mapping, ingesting from CSV/JSON/Parquet/XLSX or SQL backbones (PostgreSQL, DuckDB), or Terraform-style ontology lifecycle management. tags:

• ontology
• rdf
• owl
• sparql
• knowledge-graph
• semantic-web
• mcp
• oxigraph
• shacl
• boro metadata: openclaw: requires: mcp: - name: open-ontologies description: > Oxigraph-backed MCP server providing all onto_* tools. Install: cargo install --git https://github.com/fabio-rovai/open-ontologies OR download a binary from https://github.com/fabio-rovai/open-ontologies/releases config: command: open-ontologies args: ["serve"] bins: - open-ontologies network:
  • description: "onto_pull fetches ontologies from remote URLs or SPARQL endpoints (user-provided URLs only)" direction: outbound optional: true
  • description: "onto_push sends triples to a user-specified SPARQL endpoint" direction: outbound optional: true notes: > All processing is local by default. The in-memory Oxigraph triple store runs inside the MCP server process -- no database, no JVM, no external services required. Network access is only used by onto_pull and onto_push when the user explicitly provides a remote URL or SPARQL endpoint. onto_monitor alerts are logged locally to stdout; no external notification services are contacted. No credentials or API keys are needed for core functionality.

Open Ontologies

AI-native ontology engineering. Generate OWL/RDF directly, validate with MCP tools, iterate until clean, govern with a Terraform-style lifecycle.

Prerequisites

This skill requires the Open Ontologies MCP server to provide the onto_* tools.

Install: cargo install --git https://github.com/fabio-rovai/open-ontologies --features embeddings,plugins,sql, or download from GitHub releases, which ships the default feature set without embeddings, SQL or plugins. The crate is not on crates.io yet, so plain cargo install open-ontologies does not work

MCP config (add to .mcp.json or Claude settings):

{
  "mcpServers": {
    "open-ontologies": {
      "command": "open-ontologies",
      "args": ["serve"]
    }
  }
}

No credentials needed. All processing runs locally in an in-memory Oxigraph triple store. Network access is only used when you explicitly call onto_pull (fetch remote ontology) or onto_push (send to SPARQL endpoint) with a URL you provide. Monitor alerts (onto_monitor) are logged to stdout only.

Core Workflow

When building or modifying ontologies, follow this workflow. Decide which tools to call and in what order based on results -- this is not a fixed pipeline.

1. Generate

• Understand the domain requirements (natural language, competency questions, methodology constraints)
• Generate Turtle/OWL directly -- Claude knows OWL, RDF, BORO, 4D modeling natively
• For complex methodologies, ask for background documents or constraints

2. Validate and Load

• Call onto_validate on the generated Turtle -- if it fails, fix syntax errors and re-validate
• Call onto_load to load into the Oxigraph triple store
• Call onto_stats to verify class count, property count, triple count match expectations

3. Verify

• Call onto_lint to check for missing labels, comments, domains, ranges -- fix any issues found
• Call onto_query with SPARQL to verify structure (expected classes, subclass hierarchies, competency questions)
• If a reference ontology exists, call onto_diff to compare

4. Iterate

• If any step reveals problems, fix the Turtle and restart from step 2
• Continue until validation passes, stats match, lint is clean, and SPARQL queries return expected results

5. Persist

• Call onto_save to write the final ontology to a .ttl file
• Call onto_version to save a named snapshot for rollback

Ontology Lifecycle (Terraform-style)

For evolving ontologies in production:

1. Plan -- onto_plan shows added/removed classes, blast radius, risk score. Check onto_lock for protected IRIs.
2. Enforce -- onto_enforce with a rule pack (generic, boro, value_partition) checks design pattern compliance.
3. Apply -- onto_apply with mode safe (clear + reload) or migrate (add owl:equivalentClass bridges).
4. Monitor -- onto_monitor runs SPARQL watchers with threshold alerts. Use onto_monitor_clear if blocked.
5. Drift -- onto_drift compares versions with rename detection and self-calibrating confidence.

Data Extension Workflow

When applying an ontology to external data:

1. onto_map -- generate mapping config from data schema + loaded ontology
2. onto_ingest -- parse structured data (CSV, JSON, NDJSON, XML, YAML, XLSX, Parquet) into RDF
3. onto_shacl -- validate against SHACL shapes (cardinality, datatypes, classes)
4. onto_reason -- run RDFS or OWL-RL inference, materializing inferred triples
5. Or use onto_extend to run the full pipeline: ingest, SHACL validate, reason in one call

Clinical Terminology Support

For healthcare ontologies:

• onto_crosswalk -- look up mappings between ICD-10, SNOMED CT, and MeSH
• onto_enrich -- add skos:exactMatch triples linking classes to clinical codes
• onto_validate_clinical -- check class labels against clinical crosswalk terminology

Ontology Alignment

For aligning two ontologies:

• onto_align -- detect alignment candidates (equivalentClass, exactMatch, subClassOf) using 6 weighted signals
• onto_align_feedback -- accept/reject candidates to self-calibrate confidence weights

Tool Reference

Usage Examples

Build a pizza ontology from scratch

Build me a pizza ontology with classes for Pizza, PizzaBase (ThinAndCrispy, DeepPan),
PizzaTopping (Mozzarella, Tomato, Pepperoni, Mushroom), and properties hasBase, hasTopping.
Include rdfs:labels and rdfs:comments on everything. Validate and run competency queries
to check I can ask "what toppings does a Margherita have?"

Load and query an existing ontology

Load the ontology from https://www.w3.org/TR/owl-guide/wine.rdf, show me stats,
lint it, and run a SPARQL query to find all subclasses of Wine.

Evolve an ontology safely

I need to add a new class "GlutenFreePizza" as a subclass of Pizza with a restriction
that hasBase only GlutenFreeBase. Plan the change, enforce against generic rules,
and apply in safe mode.

Ingest CSV data into a knowledge graph

I have a CSV of employees with columns: name, department, role, start_date.
Map it to the loaded HR ontology and ingest it. Then validate with SHACL shapes
and run inference to materialize department hierarchies.

Align two ontologies

Load schema.org and my company ontology. Run onto_align to find equivalentClass
and exactMatch candidates. I'll review and give feedback to calibrate the weights.

Key Principle

Dynamically decide the next tool call based on what the previous tool returned. If onto_validate fails, fix and retry. If onto_stats shows wrong counts, regenerate. If onto_lint finds missing labels, add them. The MCP tools are individual operations -- Claude is the orchestrator.

One triple. Nothing added, nothing removed, blast radius zero. 901 consequences that were not there before.

$ printf 'load base.ttl\nplan proposed.ttl\n' | open-ontologies batch -
#   the whole change:  ex:hasParent rdfs:domain ex:Person

added_classes         0
removed_classes       0
blast_radius          0 triples affected
risk_score            low
                      ────────────────────────────────────────────
conservativity        not_conservative_under_rule_table
new consequences      901          rule table owl-rl, in 0.04s

Each number from a shape diff tells you that this change is safe. But the change gave a new type to each individual that the property already had. The tool closes this gap. A text diff cannot show you the gap. The change is one correct line, and the text diff is one line long.

Then give the reviewer the proof. The run writes a certificate. A different person re-verifies months later. That person needs no instance of this software and no network. The command is oo-cert asserted.tsv derivations.tsv. The exit code is 0, and the theorem OOCert.certificate_sound covers the result.

The checker also refuses a forged proof. Write a false conclusion into the derivation file. The same checker exits 1 and names the rule that does not hold. This is the red edge in the figure above. This refusal, and not the headline, is the part that survives examination.

This is not an ontology editor. Use Protégé to draw class hierarchies. You run Open Ontologies on the change, before the change goes to production.

The tool is like Terraform, and this is on purpose. But the plan is semantic, not syntactic. A text diff is git diff, and you have git diff already.

You do not need a JVM. You do not need Protégé. The engine speaks MCP to Claude, to Cursor, and to other clients of that protocol.

See the engine at work

Three triples go in, and three triples come out. A person asserted only that ex:Northwind is in a sanctioned jurisdiction. The engine derived the need for enhanced due diligence. A different person can check that derivation. That person does not have to trust you, or this engine, or the model that wrote the ontology.

Watch the last seconds. Somebody forges one conclusion, and leaves the two premises exactly as they were. The same checker refuses the conclusion and names the rule. oo-horn printed each line in that terminal for the fixtures in tests/fixtures/horn/supplier/. A test runs the checker again. The test fails if the figure and the checker do not agree.

With a proof, and without a proof

If the tool measures a property, the tool says measured. If a prover gives an opinion, that opinion never takes the vocabulary of the checker. Read [what the tool