Eidetic MCP server has well-structured tool definitions with complete schemas and descriptions. All 7 tools have inputSchema with typed properties and descriptions. Tool names follow verb_noun convention (memory_search, memory_reindex, export_vault). Descriptions are substantive (100-300+ chars), explaining WHAT the tool does and WHEN to use it. However, several parameters lack descriptions (e.g., memory_health and memory_lint have empty properties objects), and output schemas are not documented. Error handling guidance is minimal, tools do not explain recovery paths or categorize errors. Parameter constraints could be more explicit (e.g., polish_count max 500 is mentioned in description but not in schema). No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications in the spec.

• Add output schema documentation for all tools. For memory_search, document: {results: [{detail_id, title, confidence, lifecycle_status}], no_confident_results: boolean, drift_diagnostics: object}. This enables agents to extract detail_id for follow-up calls.
• Add tool annotations: mark memory_search, memory_search_detail, memory_serendipity, memory_health, memory_lint with readOnlyHint=true; mark memory_reindex and export_vault with destructiveHint=true. This signals to agents which tools are safe to retry.
• For memory_health and memory_lint, add parameter descriptions even though properties are empty. E.g., 'No parameters required. Call to check system health or lint memory files.' This clarifies intent.
• Add JSON Schema constraints: export_vault.polish_count should have minimum: 0, maximum: 500; timeout should have minimum: 30, maximum: 1800. Remove these from description text and let schema enforce them.
• Enhance error handling: document what memory_search returns when no_confident_results=true and guide the agent to try memory_serendipity or broaden the query. For memory_reindex, explain when a full rebuild is needed and how to detect index corruption.