This document lists all the tools supported by the Semem MCP (Model Context Protocol) server. These tools enable AI agents to interact with the semantic memory system through standardized interfaces.
Update: The MCP tool architecture has been refactored and consolidated:
- Core Tools (7 essential tools) - Implemented in
src/mcp/tools/core/ - Unified Router - Single tool routing system in
src/mcp/server/handlers/tool-router.js - Module System - Specialized tools organized in
src/mcp/tools/modules/ - Deprecated Tools - Legacy tools automatically redirected to core equivalents
Benefits: Eliminated redundancy, simplified maintenance, clear separation of concerns, and consistent tool patterns.
Core semantic memory operations following natural language patterns.
Tool Name: tell
Description: Store information in semantic memory
Parameters:
content(string, required): Information to storetype(enum, optional): Type of content ('interaction', 'document', 'concept'), default: 'interaction'metadata(object, optional): Additional metadatalazy(boolean, optional): Enable lazy processing, default: false
Tool Name: ask
Description: Query semantic memory for information
Parameters:
question(string, required): Question to askmode(enum, optional): Query mode ('basic', 'standard', 'comprehensive'), default: 'standard'useContext(boolean, optional): Use contextual search, default: trueuseHyDE(boolean, optional): Use Hypothetical Document Embeddings, default: falseuseWikipedia(boolean, optional): Include Wikipedia search, default: falseuseWikidata(boolean, optional): Include Wikidata search, default: falseuseWebSearch(boolean, optional): Include web search, default: false
Tool Name: augment
Description: Enhance stored information with additional processing
Parameters:
target(string, optional): Target to augment, default: 'all'operation(enum, optional): Type of operation ('auto', 'concepts', 'attributes', 'relationships', 'process_lazy', 'chunk_documents', 'extract_concepts', 'generate_embedding', 'analyze_text', 'concept_embeddings'), default: 'auto'options(object, optional): Configuration options for chunking, concept extraction, and augmentation
Spatial metaphor navigation through knowledge graphs using Zoom/Pan/Tilt operations.
Tool Name: zoom
Description: Change the granularity level of view
Parameters:
level(enum): Zoom level ('entity', 'concept', 'document', 'community'), default: 'entity'query(string, optional): Focus query
Tool Name: pan
Description: Navigate across different domains or relationships
Parameters:
direction(enum, optional): Pan direction ('semantic', 'temporal', 'conceptual')domain(string, optional): Domain to pan totimeRange(string, optional): Time range filterconceptFilter(array, optional): Concept filterssemanticVector(array, optional): Semantic direction vectormaxResults(number, optional): Maximum results, default: 50threshold(number, optional): Similarity threshold, default: 0.5
Tool Name: tilt
Description: Change the perspective or representation style
Parameters:
style(enum): View style ('keywords', 'summary', 'detailed'), default: 'keywords'query(string, optional): Focus query
Tool Name: inspect
Description: Examine system state or specific components
Parameters:
type(enum): Inspection type ('system', 'session', 'concept', 'memory'), default: 'system'target(string, optional): Specific target to inspectincludeRecommendations(boolean, optional): Include recommendations, default: false
Advanced memory operations for persistent storage and retrieval.
Tool Name: remember
Description: Store information with importance and context metadata
Parameters:
content(string, required): Content to rememberimportance(enum, optional): Importance level ('low', 'medium', 'high', 'critical'), default: 'medium'domain(string, optional): Domain classificationtags(array, optional): Tags for categorizationcontext(string, optional): Contextual informationmetadata(object, optional): Additional metadata (source, timestamp, category)
Tool Name: forget
Description: Remove or fade information from memory
Parameters:
target(string, required): Target to forgetmethod(enum, optional): Forgetting method ('fade', 'remove'), default: 'fade'intensity(number, optional): Intensity of forgetting (0-1), default: 0.5
Tool Name: recall
Description: Retrieve stored memories based on criteria
Parameters:
query(string, required): Recall querydomain(string, optional): Domain filtertimeRange(string, optional): Time range filtertags(array, optional): Tag filterslimit(number, optional): Maximum results, default: 10threshold(number, optional): Similarity threshold, default: 0.5
Tool Name: project_context
Description: Project context into different dimensional spaces
Parameters:
query(string, required): Context queryprojectionType(enum): Projection type ('semantic', 'temporal', 'conceptual'), default: 'semantic'dimensions(number, optional): Number of dimensions, default: 50includeMetadata(boolean, optional): Include metadata, default: truetimeWindow(string, optional): Time windowconceptFilters(array, optional): Concept filters
Tool Name: fade_memory
Description: Gradually fade memories over time
Parameters:
target(string, optional): Specific target to fadedomain(string, optional): Domain to fadefadeRate(number, optional): Rate of fading (0-1), default: 0.1preserveImportant(boolean, optional): Preserve important memories, default: truedryRun(boolean, optional): Preview without applying, default: false
Direct SPARQL operations for advanced knowledge graph manipulation.
Description: Execute SPARQL SELECT or ASK queries Parameters:
query(string, required): SPARQL query to executelimit(number, optional): Maximum results, default: 100timeout(number, optional): Timeout in milliseconds, default: 30000
Description: Execute SPARQL CONSTRUCT queries to generate RDF Parameters:
query(string, required): SPARQL CONSTRUCT queryformat(enum, optional): Output format ('turtle', 'n-triples', 'json-ld', 'rdf-xml'), default: 'turtle'timeout(number, optional): Timeout in milliseconds, default: 30000
Description: Navigate knowledge graph using ZPT spatial metaphors Parameters:
zoom(enum, required): Zoom level ('micro', 'entity', 'relationship', 'community', 'corpus')filters(object, optional): Filter criteria (type, label, property, value, dateRange)limit(number, optional): Maximum results, default: 50orderBy(enum, optional): Result ordering ('similarity', 'frequency', 'centrality', 'weight', 'timestamp')
Description: Perform vector similarity search using embeddings Parameters:
embedding(array, required): Query embedding vector (100-2000 dimensions)threshold(number, optional): Similarity threshold (0-1), default: 0.7limit(number, optional): Maximum results, default: 10entityTypes(array, optional): Filter by entity typesincludeMetadata(boolean, optional): Include metadata, default: true
Description: Validate knowledge graph integrity and health Parameters:
checkIntegrity(boolean, optional): Check graph integrity, default: truecheckEmbeddings(boolean, optional): Check embedding coverage, default: truecheckConnectivity(boolean, optional): Check graph connectivity, default: truedetailed(boolean, optional): Return detailed report, default: false
Description: Store RDF dataset in knowledge graph Parameters:
rdfData(string, required): RDF data to storeformat(enum, optional): Input format ('turtle', 'n-triples', 'json-ld', 'rdf-xml'), default: 'turtle'graphName(string, optional): Named graph for storagereplace(boolean, optional): Replace existing data, default: false
Description: Execute multiple SPARQL operations in batch Parameters:
operations(array, required): Operations to execute (type: 'insert', 'update', 'delete')atomic(boolean, optional): Execute as atomic transaction, default: truecontinueOnError(boolean, optional): Continue on errors, default: false
Description: Manage named graphs (create, delete, clear, copy, move) Parameters:
operation(enum, required): Operation type ('create', 'delete', 'clear', 'copy', 'move')graphName(string, required): Target graph namesourceGraph(string, optional): Source graph (for copy/move)force(boolean, optional): Force operation, default: false
Knowledge graph decomposition and corpus analysis tools.
Description: Decompose text corpus into semantic units, entities, and relationships Parameters:
- Text processing and decomposition parameters
- RDF export options
- SPARQL endpoint configuration
Description: Enrich decomposed knowledge graph with vector embeddings Parameters:
- Embedding model configuration
- Batch processing options
Description: Augment entities with additional attributes and properties Parameters:
- Attribute extraction configuration
- Entity enhancement options
Description: Identify and aggregate entity communities in knowledge graph Parameters:
- Community detection algorithms
- Aggregation criteria
Vector Self-Organizing Map tools for clustering and visualization.
Description: Create new VSOM instance for clustering Parameters:
mapSize(array, optional): Grid dimensions [width, height], default: [20, 20]topology(enum, optional): Grid topology ('rectangular', 'hexagonal'), default: 'rectangular'embeddingDimension(number, optional): Vector dimension, default: 1536maxIterations(number, optional): Training iterations, default: 1000- Training and clustering parameters
Description: Load data into VSOM instance Parameters:
instanceId(string, required): VSOM instance IDdataType(enum, required): Data type ('entities', 'sparql', 'vectorIndex', 'sample')data(object, required): Data configuration based on type
Description: Train VSOM with loaded data Parameters:
- VSOM instance configuration
- Training parameters
Description: Extract clusters from trained VSOM Parameters:
- Clustering parameters
- Output format options
High-level research and learning workflow tools.
Description: Capture and process user feedback for system improvement Parameters:
query(string, required): Original queryresponse(string, required): System responsefeedback(object, required): User feedback (rating, relevance, comments)userContext(object, optional): User context information
Description: Perform incremental learning from feedback data Parameters:
learningData(array, required): Learning examples with queries and responsesoptions(object, optional): Learning configuration options
The following tools have been deprecated and are automatically redirected to core tools:
semem_extract_concepts→augmentwithoperation: 'extract_concepts'semem_generate_embedding→augmentwithoperation: 'generate_embedding'
semem_store_interaction→tellwith interaction formatuploadDocument→tellwith document processing
semem_retrieve_memories→askwith memory retrievalsemem_answer→askwith comprehensive modesemem_ask→askwith standard processing
remember→augmentwithoperation: 'remember'forget→augmentwithoperation: 'forget'fade_memory→augmentwithoperation: 'fade'recall→askwith memory domainproject_context→augmentwithoperation: 'project_context'
Note: All deprecated tools continue to work seamlessly through automatic redirection to the appropriate core tools.
- All tools return JSON responses with standardized structure including success status, results, and metadata
- Error handling is implemented consistently across all tools
- Tools support both synchronous and asynchronous operations
- Authentication and authorization are handled at the MCP server level
- Tools can be chained together for complex workflows
- All operations are logged for debugging and monitoring purposes
Tools are configured through the main MCP server configuration. See the main documentation for setup instructions and configuration options.