Server has 7 tools with mostly complete schemas and descriptions. Naming follows verb_noun convention (read_*, create_*, add_*, update_*, ask_). Descriptions are present but brief (avg ~60 chars, below the 194-char baseline). Input schemas are well-formed with types and required fields. However, output schemas are not documented, LLMs cannot infer what fields to expect from responses. Error handling and recovery guidance are absent. No parameter descriptions for complex nested objects (e.g., 'themes' array in create_themes lacks detail on theme object structure). ask_user tool is unusual, it blocks execution waiting for user input, which violates stateless request handling in current MCP spec.

• Document output schemas for all tools. Example: read_map should return {revision: number, nodes: [{id, label, detail, status, themeId}], themes: [{id, name, color}]}. This enables LLMs to chain calls correctly.
• Expand create_themes and add_nodes descriptions to include the structure of theme/node objects. Use JSON Schema examples or inline descriptions: 'themes: array of {name: string, color: string (hex), description?: string}'.
• Replace ask_user blocking behavior with Multi-Round-Trip Requests. Return {status: 'input_required', questions: [...]} and let the client handle user interaction asynchronously, then resume with the answers.
• Add error handling guidance to tool descriptions. E.g., update_node: 'Returns error if node ID not found, call read_map to list valid node IDs. Idempotent: safe to retry with same requestId.'
• Increase tool descriptions to 100 - 200 chars. Include dependency hints: 'Call read_map first to discover available nodes and themes before creating new ones.'