Loading ToolBench

Reading the current corpus…

neo4j-knowledge-mcp MCP server · F · ToolBench

← Browse

neo4j-knowledge-mcp

github.com/ryaker/neo4j-knowledge-mcp

Neo4j Knowledge Graph MCP Server for sophisticated knowledge management

2026-07-28+STDIO8 tools

39/100

Definition Quality48

Protocol Readiness35

Supportability28

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence

Current-spec patterns detected

Summary

Server has 8 tools with schemas and descriptions present, but quality is inconsistent. Tool naming lacks action verbs (e.g., 'store-knowledge' vs 'store_knowledge'; 'explore-knowledge-graph' is compound). Descriptions range 34 - 194 chars but lack WHEN/WHY guidance. Parameter descriptions exist but are generic ('Additional metadata', 'Filters for domain'). Output schemas are undocumented, no response structure defined. Error handling is minimal (generic -32603 errors). Tools 3 - 7 are inferred from index.js references without explicit registration visible in mcp-server.js, capping their scores at 50. Overall: below baseline for production use.

Tools (8)

aboutread onlysource verified75/100

Get information about the Neo4j Knowledge Graph MCP server

analyze-knowledge-gapsread onlysource verified38/100

Identify gaps in the knowledge graph for targeted research

explore-knowledge-graphread onlysource verified30/100

Interactively explore the knowledge graph from a starting concept

find-knowledge-pathsread onlysource verified35/100

Find connection paths between concepts in the knowledge graph

process-mcp-outputwritesource verified35/100

Process outputs from other MCP servers into the knowledge graph

search-knowledgeread onlysource verified68/100

Search the knowledge graph using semantic and graph-based queries

store-document-knowledgewritesource verified37/100

Extract and store knowledge from documents processed by other MCP servers

Issues (5)

CRITICALtool-registration ↗

Tools 3 - 7 (explore-knowledge-graph, find-knowledge-paths, process-mcp-output, analyze-knowledge-gaps, store-document-knowledge) are referenced in index.js but NOT explicitly registered in mcp-server.js tools/list response. Only 3 tools (store-knowledge, search-knowledge, about) are actually exposed. This violates tool registration and makes 5 tools unreachable.

explore-knowledge-graphfind-knowledge-pathsprocess-mcp-outputanalyze-knowledge-gapsstore-document-knowledge

HIGHresponse-shaper ↗

No output schemas documented for any tool. LLMs cannot plan downstream calls or extract structured data. E.g., store-knowledge returns {success, id, error} but this is inferred, not declared. search-knowledge response structure is unknown.

store-knowledgesearch-knowledgeexplore-knowledge-graphfind-knowledge-pathsprocess-mcp-outputanalyze-knowledge-gapsstore-document-knowledge

Recommendations

Register ALL 8 tools in mcp-server.js tools/list response. Currently only 3 are exposed. Add handlers for explore-knowledge-graph, find-knowledge-paths, process-mcp-output, analyze-knowledge-gaps, store-document-knowledge in the tools/call switch statement.
Document output schemas for every tool. E.g., store-knowledge returns {success: boolean, id: string, error?: string}; search-knowledge returns {results: [{id, content, score, source}], total: number, nextCursor?: string}. Include these in tool definitions and in handler comments.
Rename tools to use underscores and action verbs: store_knowledge, search_knowledge, explore_knowledge_graph → explore_graph, find_knowledge_paths → find_paths, process_mcp_output → extract_and_store_mcp_output, analyze_knowledge_gaps → identify_knowledge_gaps, store_document_knowledge → extract_and_store_document_knowledge.
Add enum constraints and format descriptions to parameters. E.g., searchType description: 'Type of search: semantic (vector similarity), graph (Cypher traversal), hybrid (both), or exact (string match).' For contextFilters, specify: 'Object with optional keys: domain (string), source (string), confidence (0.0 - 1.0).'
Implement error classification. Return {error: {code, message, recoveryHint, retryable: boolean}}. E.g., 'Neo4j connection timeout. Retryable: true. Hint: Check NEO4J_URI and network connectivity.' For invalid enums: 'Invalid searchType: got "fuzzy". Must be one of: semantic, graph, hybrid, exact.'

Score history

Overall score trend

First recorded score · v2 rubric

39/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	39	2026-07-28+	v2