
Open Ontologies MCP Server: Setup for Claude, Cursor and Other Apps
Lets Claude plan, validate and reason over changes to an OWL ontology, and give reviewers a proof they can re-check.
Open Ontologies is a community server for working with OWL and RDF ontologies. It lets your assistant load an ontology, check it, reason over it and preview a change with a risk score before you apply it. It can also write a certificate that a separate checker can verify.
You download one binary and start it with its serve command. It runs on your computer and needs no account or key. It has more than 100 tools, and the list below shows the main ones.
Good for
- Ontology and knowledge-graph engineers who change production vocabularies
- Reviewers who need evidence for each ontology change
- Data teams who turn spreadsheets and SQL data into RDF
Works in
12 should work
Tap an app to see how to set it up and any known problems.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Supports local MCP servers.
Client does not support this transport.
Client does not support this transport.
Install Open Ontologies MCP
Claude Code
- 1
Copy this prompt.
Add the Open Ontologies MCP server to Claude Code. Run this command: claude mcp add open-ontologies -- open-ontologies serve Then check that it loaded and tell me if I need to sign in or restart.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires open-ontologies binary on your PATH (download from the GitHub releases).
Download the release binary for your system first, as the README's Install section shows, and put it on your PATH. Run open-ontologies init once to create its config and database. The Windows steps are in the project's Windows guide.
Built from the package data we hold. Not yet tested by us.
Tools
Previews what a change to an ontology would do
Compares the loaded ontology with proposed Turtle, like a Terraform plan. It reports what is added or removed, how many triples a change touches and a risk score, before anything is applied.
Try: “Plan this change to the ontology and tell me the blast radius.”
Read ·
onto_planApplies a previewed change
Applies the plan you made earlier, the most recent one unless you name another. Use it after you have read the plan.
Try: “Apply the plan we just reviewed.”
Write ·
onto_applyProtects names from being removed
Locks IRIs so that a plan or apply cannot remove them. Use it for terms other systems depend on.
Try: “Lock the IRIs for Person and Organisation.”
Write ·
onto_lockSpots changes between two versions
Compares two versions of an ontology and reports what drifted between them.
Try: “Has the ontology drifted since the last release?”
Read ·
onto_driftRuns watchers that can block changes
Runs the active monitoring watchers over the ontology. If one trips, applying changes is blocked until you clear it.
Try: “Run the monitors and tell me if anything is blocked.”
Read ·
onto_monitorRestores an earlier saved version
Puts the in-memory ontology back to a version you saved with the snapshot tool.
Try: “Roll back to the snapshot called before-merge.”
Write ·
onto_rollbackSaves a named snapshot
Stores a named snapshot of the current ontology so you can return to it later.
Try: “Save a snapshot called before-merge.”
Write ·
onto_versionLists saved snapshots
Shows the saved ontology versions.
Try: “What snapshots do I have?”
Read ·
onto_historyLoads an ontology from a file or pasted Turtle
Reads an ontology file, or Turtle text you give it, into the server's working store. Most other tools work on what is loaded.
Try: “Load pizza.ttl.”
Files ·
onto_loadChecks that ontology syntax is valid
Checks RDF or OWL syntax in a file or in text and reports parse errors.
Try: “Validate this Turtle file.”
Read ·
onto_validateRuns a SPARQL query
Runs a SPARQL query against the loaded ontology and returns the rows.
Try: “List every class and its parent.”
Read ·
onto_queryCounts triples, classes and properties
Reports how many triples, classes, properties and individuals are loaded.
Try: “Give me stats on the loaded ontology.”
Read ·
onto_statsFinds quality problems
Checks the ontology for missing labels, comments, domains and ranges.
Try: “Lint the ontology and list what is missing.”
Read ·
onto_lintCompares two ontology files
Shows which triples were added or removed between two ontology files.
Try: “What changed between v1.ttl and v2.ttl?”
Files ·
onto_diffDerives what follows from the ontology
Runs inference over the loaded ontology and adds the conclusions it can derive. It can also write a certificate that a separate checker can verify.
Try: “Run the reasoner and write a certificate.”
Write ·
onto_reasonChecks data against SHACL shapes
Validates the loaded data against SHACL shapes and reports the failures.
Try: “Check the data against these shapes.”
Read ·
onto_shaclTests whether one class is a subclass of another
Uses description-logic reasoning to answer whether one class is subsumed by another.
Try: “Is Margherita a kind of Pizza?”
Read ·
onto_dl_checkExplains why a class can never have members
Gives the reason a class is unsatisfiable under description-logic reasoning.
Try: “Why is VeganCheesePizza unsatisfiable?”
Read ·
onto_dl_explainFinds which asserted facts cause a conclusion
Names the asserted triples responsible for a conclusion or a contradiction.
Try: “Which facts make this class inconsistent?”
Read ·
onto_justifyChecks that new axioms change nothing old
Reasons over the base ontology with and without an extension and reports new conclusions about names the base already used.
Try: “Does adding this module change anything the base said?”
Read ·
onto_conservative_checkTurns a data file into RDF
Parses CSV, JSON, NDJSON, XML, YAML, XLSX or Parquet into triples and loads them.
Try: “Load customers.csv as RDF.”
Write ·
onto_ingestBuilds an ontology from one spreadsheet
Takes one data sheet and produces an OWL class, typed properties, a SHACL shape and a loading mapping, with the evidence for each line.
Try: “Make an ontology from this spreadsheet.”
Files ·
onto_induceLoads database rows as RDF
Runs a SQL query against PostgreSQL or DuckDB and loads the rows as triples. It needs a build with that option, so the standard binaries do not have it.
Try: “Load this query's rows into the ontology.”
Network ·
onto_sql_ingestFinds matching terms in two ontologies
Looks for equivalent or related classes between two ontologies using labels, properties, parents and neighbours.
Try: “Align these two vocabularies.”
Read ·
onto_alignFetches an ontology from a URL
Downloads an ontology from a remote URL or SPARQL endpoint and loads it.
Try: “Pull the ontology from this URL.”
Write ·
onto_pullSends the ontology to a SPARQL endpoint
Sends the whole current ontology store to a SPARQL endpoint you name, as an INSERT DATA update. The call can carry a username, password or token for that endpoint.
Try: “Push the current ontology to the staging endpoint.”
Write ·
onto_pushWrites the ontology to a file
Saves the current ontology store to a file you name.
Try: “Save the ontology to out.ttl.”
Write ·
onto_saveEmpties the loaded ontology
Clears all triples from the in-memory store. The compile cache file is kept.
Try: “Clear the loaded ontology.”
Write ·
onto_clearDeletes a cached ontology copy
Deletes a named entry from the compile cache. By default it also deletes the compiled N-Triples copy in the cache folder (~/.open-ontologies/cache). The original ontology file you loaded is not touched, and passing delete_file=false keeps the cache file.
Try: “Remove the pizza ontology from the cache.”
Delete ·
onto_cache_removeBrowses and installs standard ontologies
Browses the built-in catalogue of 33 standard ontologies and the community pack registry, which it fetches at run time, and installs one.
Try: “Install the FOAF ontology from the marketplace.”
Write ·
onto_marketplaceHands the problem to a theorem prover
Runs a first-order prover, reads the derivation it prints and re-checks what can be re-checked.
Try: “Prove this with Vampire and check the result.”
Runs code ·
onto_fol_proveRuns a PDDL planner
Starts Fast Downward as a separate process on a PDDL domain and problem.
Try: “Plan these steps with Fast Downward.”
Runs code ·
onto_plan_classicalRuns a tool from an installed plugin
Calls a tool in an installed WASM plugin. It needs a build with plugins on, so the standard binaries do not have it.
Try: “Call the plugin's export tool.”
Runs code ·
onto_plugin_call
Read from the source code
Permissions
medium risk · 4 things to watch
Medium risk
The assistant can load, change and save ontologies on your computer, push to remote SPARQL endpoints and start helper programs. Approve each write, and use a tool profile or deny list to keep the surface small.
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only modeHow to turn it on
Watch out for
onto_push sends your ontology to a remote endpoint
It uploads the full current store as a SPARQL INSERT DATA update to any endpoint URL in the call, and the call can carry a username, password or token for it. It adds data and does not drop graphs. Check the address before you approve it. github.com
Some tools write files and delete cached copies
onto_save, onto_pack and conversion tools write files you name. onto_cache_remove (and onto_unload with delete_cache) deletes the compiled N-Triples copy in ~/.open-ontologies/cache by default. It does not delete the source file you loaded, and reloading rebuilds the cache. Approve each call that names a path. github.com
Some tools start other programs
The planner tool starts Fast Downward and the prover tool runs a first-order prover. Both need those programs installed. github.com
The tool list is very large
The default profile exposes every tool. Use --tool-profile with a job name such as validation or authoring to expose fewer. github.com
Security advisories
No published security advisories as of Nov 1, 2026.
Technical details
| Turning on read-only | Start serve with --tools-allow @read_only to list only the read-only group (status, validate, query, stats, diff, lint and a few more), or pick a --tool-profile. |
|---|---|
| Talks to | Remote ontology URLs and SPARQL endpoints you name. Community ontology registry (fetched when you browse the marketplace) |
| Sign-in and keys | No credentials are used. Database tools take connection details you pass in a call. |
| Scopes |
|
Questions
Is Open Ontologies official?
No. It is a community project by fabio-rovai, MIT licensed. The author's company, Tesseract Semantics, sells a hosted platform on top of it.
Do I need an account or API key?
No. You download the binary and run it locally. Nothing in the README asks for a key.
Does this page list every tool?
No. The list shows the main ones. The project's own tool reference counts 122 in the source, and the standard binaries leave out 8 that need extra build options.
Can I run it read-only?
Partly. You can start it with only the read-only group of tools, or with a smaller profile. The profiles are built for jobs, so they still include tools that write.
Is it an ontology editor like Protege?
No. The README says it is not an editor. It checks and previews changes and proves what follows from them.
Does it need Java?
No. The README says you need neither a JVM nor Protege.
How do I add it to Claude Code or Cursor?
Download the binary as the README shows, then use your client's tab in the install box above.
Information
updated Nov 2, 2026
- Publisher
- Community
- Transport
- stdio
- Auth
- None
- License
- MIT
- Last commit
- Oct 30, 2026
- Open issues
- 3
Sources
Docs, issue trackers and discussions this page is built from. Issues, quotes and findings above link to their own source.
Show all 4Hide
- github.com · github.com
- github.com · github.com
- github.com · github.com
- github.com · github.com