Tools

All 105 tools the MCP server exposes, grouped by task. Each category page documents every tool with its arguments and returns.

Version 0.8.1 adds the general write_project_policy updater to the 104-tool v0.8.0 surface. The server uses the latest Java MCP SDK and MCP 2025-11-25 tool contract: standard boolean annotations classify read-only, destructive, idempotent, and open-world behavior, while each tool and input-property description carries its exact workflow, argument mapping, and recovery guidance.

New tool in 0.8.1
New tools in 0.8.0
New tool in 0.7.2
New tool in 0.7.1
New tools in 0.7.0
New tools in 0.6.0
Conventions
A typical flow
Tool index

New tool in 0.8.1

Task	New tool	What it adds
Project policy & metadata	`write_project_policy`	Scaffold a valid commented `.protege-mcp/project.yaml` with RO-Crate metadata, replace complete authored YAML, or recursively patch policy fields while preserving unaffected content and upgrading v1/v2 to v3.

New tools in 0.8.0

Task	New tools	What they add
SSSOM mappings	`list_mappings`, `add_mapping`, `remove_mapping`, `import_sssom`, `export_sssom`, `validate_mappings`	Governed SSSOM 1.0 authoring with deterministic IDs, project validation, revision-bound pagination, compare-and-swap writes, and atomic import/export.
External terms	`search_external_terms`, `inspect_external_term`, `propose_term_reuse`, `accept_reuse_proposal`	Project-governed OLS4 plus experimental OntoPortal discovery with BioPortal/AgroPortal endpoint presets, immutable revision-bound reuse proposals, and explicit receipt/CAS/saga acceptance.
Reasoner and rules	`get_reasoner_capabilities`, `validate_rules`, `materialize_inferences`, `commit_materialization`	Exact capability evidence, non-executing SWRL validation, and preview-first bounded inference publication.
Asynchronous jobs	`start_job`, `get_job`, `cancel_job`, `list_jobs`, `export_job_artifact`	Owner-scoped live jobs with immutable captures, bounded execution, monotonic cancellation, private verified artifacts, and guarded atomic export.

New tool in 0.7.2

Task	New tool	What it adds
Ontology metadata	`remove_prefix`	Remove exactly one prefix name while preserving same-namespace siblings and standard prefixes; like `set_prefix`, the document-format edit is immediate and not undoable.

New tool in 0.7.1

Task	New tool	What it adds
Audit	`export_audit_log`	Deterministically merge the active project’s rotated, owner-only streams into one bounded, re-redacted JSONL artifact. Dry-run is the default; a write requires a valid project policy, server administration plus project read/write capabilities, and the normal confirmation gate.

New tools in 0.7.0

Version 0.7.0 adds 5 tools to the 78-tool v0.6.x surface. Existing tool arguments and default interactive behavior remain compatible.

Task	New tools	What they add
Transactional editing	`rebase_change_set`	Deterministically re-resolve a cached preview at the current revision; a changed resolution fails closed for human review.
Change review	`analyze_change_impact`	Syntactic impact analysis of a cached change set or asserted diff: affected entities and modules, referencing axioms, downstream terms, foreign re-axiomatization, deprecated terms in use, and validation assets naming changed IRIs.
Release	`run_release_gate`, `prepare_release`	Run the strict QC gate plus the release-only checks (import provenance, version IRI, verified serialization round trip, fingerprint stability, optional baseline diff) read-only, then produce the manifest, reports, and RO-Crate bundle — dry-run by default, written atomically into the policy output directory on confirmation.
Project policy & QC	`write_project_policy_template`	Scaffold an immediately valid commented `.protege-mcp/project.yaml` from the active ontology.

New tools in 0.6.0

Version 0.6.0 adds 12 tools to the 66-tool v0.5.x surface. Existing tool arguments and default interactive behavior remain compatible.

Task	New tools	What they add
Project policy & QC	`get_project_policy`, `validate_project_policy`, `run_project_qc`	Discover a checked-in policy, validate its files and settings, and run its required checks as one reproducible gate.
Revisions & release evidence	`get_model_revision`, `semantic_diff`	Pin the exact workspace revision used by a change and classify asserted changes between ontology artifacts.
Transactional editing	`preview_change_set`, `commit_change_set`, `discard_change_set`	Preview an exact edit against an isolated QC snapshot, commit only if the revision still matches, or discard it.
Import integrity	`inspect_imports`, `write_import_lock`, `verify_import_lock`, `validate_catalog`	Inspect the loaded import graph, lock local dependencies by checksum, verify the lock offline, and validate an OASIS catalog.

The save_ontology, load_ontology, merge_ontology_document, create_terms, create_properties, and run_qc_suite tools also gain optional v0.6.0 behavior; their existing direct-call defaults are unchanged. See each tool’s section for its current arguments and result contract.

Conventions

These hold for every tool:

MCP-native behavior and guidance. Every tool publishes complete standard ToolAnnotations, defined alongside its description and input schema in the single mcp-catalog.json source of truth. Because those fields are boolean hints rather than procedural instructions, sequencing, exact argument names, non-invention rules, and failure recovery live in the model-visible tool description and the descriptions of the relevant input-schema properties.
Structured JSON output. Every tool returns a structured JSON object, delivered as MCP structuredContent and mirrored as a JSON text block — so every client, and a human reading the transcript, see the same result.
Entities by IRI or name. You may reference an entity by its full IRI or by its display name (the active renderer). Where noted, class/data-range operands also accept Manchester syntax expressions (e.g. hasPart some Cell).
Edits are GUI-visible and undoable. Every model edit goes through OWLModelManager.applyChanges, so it appears in Protégé immediately and joins the shared Edit ▸ Undo stack. (Document open/save operations use Protégé’s own load/save APIs and are not undoable.)
Safety gates. Mutating tools obey the read-only and confirm-each-write preferences (Settings ▸ MCP). Every tool also checks the authenticated principal’s declared ontology/release/ filesystem/network capabilities before its handler runs. Read/query tools work in global read-only mode, but a caller-selected file still needs its explicit filesystem capability.
Minting signal. Many create/add tools report new_entities (the entities the call introduced) and accept strict — set it to fail instead of minting a brand-new entity from an unrecognized IRI/name (a typo guard).
Active ontology + imports. Tools act on the active ontology (the current edit target). Where a tool accepts include_imports, it can widen to the imports closure.

A typical flow

A safe natural-language editing loop:

Orient (get_ontology_context / get_entity_context) → ground a name to an IRI, or confirm it would mint a new one (search_entities — would_mint / best_match) → preview an edit (preview_changes) → apply & verify in one call (apply_changes with verify=rollback prevents a delta that fails the effective change-set gate) → gate with the requirements suite and invariants (run_competency_questions, verify_ontology, or the umbrella run_qc_suite).

The top-level Prompts guide packages these flows for one-click use in an MCP client.

Tool index

Explore & search

list_ontologies · get_active_ontology · summarize_ontology · list_classes · search_entities · get_entity · get_axioms_for_entity

Context & validation

get_ontology_context · get_entity_context · get_model_revision · validate_ontology · validate_governance · diff_ontologies · semantic_diff · analyze_change_impact

Safe authoring & QC

get_project_policy · validate_project_policy · run_project_qc · write_project_policy_template · write_project_policy · run_release_gate · prepare_release · export_audit_log · verify_ontology · run_qc_suite · shacl_validate · add_competency_question · list_competency_questions · remove_competency_question · run_competency_questions (plus apply_changes verify= and search_entities grounding — see their category pages)

Editing — entities & axioms

preview_changes · apply_changes · preview_change_set · commit_change_set · discard_change_set · rebase_change_set · create_class · create_entity · create_term · create_terms · create_property · create_properties · add_subclass_of · add_annotation · set_label · add_axiom · remove_axiom · rename_entity · delete_entity · deprecate_entity · move_class

Ontology metadata & imports

set_ontology_id · set_prefix · remove_prefix · add_import · remove_import · add_ontology_annotation · remove_ontology_annotation

Documents

load_ontology · set_active_ontology · merge_ontology_document · create_ontology · inspect_imports · write_import_lock · verify_import_lock · validate_catalog · write_catalog · extract_module

Rules (SWRL)

list_rules, add_rule, remove_rule, validate_rules

Reasoning

get_reasoner_capabilities, validate_rules, materialize_inferences, commit_materialization, list_reasoners, set_reasoner, run_reasoner, get_unsatisfiable_classes, get_inferred_superclasses · execute_dl_query · explain_entailment · get_explanations · explain_inconsistency

Cross-cutting reference

Axiom types — the structured axiom_type operand catalog used by add_axiom, remove_axiom, preview_changes, apply_changes, explain_entailment, and get_explanations.
Prompts — 11 guided workflows that compose the tools into safe, repeatable tasks.