Graph Query MCP has 7 well-named, read-only tools with consistent parameter schemas and reasonable descriptions. All tools follow verb_noun naming (search_, get_, find_, query_). Descriptions are present (avg ~80 chars) and explain purpose clearly. Input schemas are complete with types and defaults. However, descriptions lack actionable context (when to use vs similar tools, prerequisites, expected output structure). Output schemas are not documented, LLMs cannot predict response structure for chaining. Error handling is minimal; no recovery guidance. Parameters accept free-form strings where enums would prevent hallucination (e.g., scope should be enum ['primary', 'secondary']). Missing per-parameter descriptions in some cases (e.g., 'limit' lacks guidance on why 20 vs 30 vs 40). No tool annotations (readOnlyHint present in metadata but not exposed in schema).

• Document output schema for each tool. Example: 'Returns {nodes: [{id, name, degree, metadata}], total_count, next_cursor}'. This enables LLMs to chain tools and extract fields correctly.
• Convert scope parameter to enum: {type: 'string', enum: ['primary', 'secondary'], description: 'Graph scope...'}. Prevents invalid values.
• Expand tool descriptions to 100-150 chars. Add: (1) When to use vs similar tools, (2) What fields are returned, (3) Any prerequisites. Example: 'Find nodes by fuzzy name match. Returns node IDs, names, and degree. Use before get_neighbors to locate a starting node. Scope: primary (full graph) or secondary (subset).'
• Add per-parameter descriptions for 'limit', 'max_hops', 'query'. Explain ranges, defaults, and side effects. Example: 'limit (1-100, default 20): Maximum results. Larger limits increase latency; use pagination for large result sets.'
• Implement error responses with recovery hints. Example: 'Node "xyz" not found in primary scope. Try search_nodes("xyz") to find similar nodes, or switch to secondary scope.'