This server has serious quality gaps across naming, descriptions, and schema completeness. Of 14 tools, 6 have Chinese descriptions that are overly verbose (>400 chars) and mixed with implementation details rather than agent-optimized guidance. 8 have sparse English descriptions (20-50 chars) that lack context for when/why to use them. Parameter schemas are present but descriptions are generic or absent. Output schemas are entirely undocumented. Error handling is not visible in the provided code. The tool names follow verb_noun convention acceptably, but the overall composition has issues: tools like create_entities and create_relations expect highly domain-specific array structures without validation guidance. No tool annotations (readOnlyHint, destructiveHint) are present. The server appears designed for a specific narrative-writing knowledge graph domain, but the interface does not translate well to general LLM-agent usage patterns.

• Rewrite create_entities and create_relations descriptions to be <100 characters in English, stating the purpose (e.g., 'Create new entities in the knowledge graph from narrative data') and when to use them. Move domain-specific format guidance (observation structure, property examples) into a separate schema docs or OpenAPI spec.
• Add documented output schemas for all tools. For create_entities, specify: {entity_id: string, name: string, entityType: string}. For read_graph, specify: {entities: [{name, type, aliases, observations}], relations: [{from, to, type, properties}]}. LLMs need field lists to chain calls.
• Add tool annotations: set destructiveHint=true on delete_* tools and readOnlyHint=true on search_*, read_*, get_* tools. This is a lightweight MCP extension (spec 2026-07-28 supports it) that eliminates LLM confusion.
• Add error handling documentation: specify which errors are retryable (e.g., temporary graph lock) vs. permanent (e.g., entity not found). For search_nodes with no results, add recovery text: 'No entities matched. Try a broader query or use read_graph to discover available entities.'
• Standardize parameter descriptions. For all 'path' params, document: 'Path format: /v<volume>/a<act>/p<plotpoint>/c<chapter>, e.g., /v1/a1/p1/c1. Or use chapter index directly: c51 or 51.' For windowSize, add: '1-10 recommended; defaults to 3 (±3 chapters).'
• Add validation constraints to create_entities and create_relations. Define entityType as an enum: ['Person', 'Item', 'Organization', 'WorldSetting', 'PowerSystem', 'Event']. For observations, provide a schema with allowed property keys (age, gender, background, etc.) instead of free-form strings.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry