Server has 5 tools with complete input schemas and descriptions. Naming follows verb_noun convention (list_, search_, get_, query_). Descriptions are present (avg ~120 chars) but lack actionable context for LLM selection. Parameters are typed with enums and constraints (e.g., collection enum, limit bounds). However, output schemas are not documented, LLMs cannot predict response structure. Error handling is absent from visible code. No tool annotations (readOnlyHint, destructiveHint). Parameter descriptions are minimal (10-30 chars) and lack dependency hints or format guidance.

• Document output schemas for all tools. Add a 'returns' section to each tool definition specifying field names, types, and meanings. Example: 'Returns {numberMatched: number, results: [{id: string, title: string, queryable: boolean, ...}]}'.
• Expand parameter descriptions to 50-100 chars with format hints and dependency notes. E.g., 'Text search query (optional, 1-500 chars; if omitted, returns all datasets in collection)' and 'Dataset ID (obtain from search_datasets result)'.
• Add readOnlyHint: true to all tool definitions in the inputSchema to signal to agents that these tools are safe to retry and do not mutate state.
• Document mutual exclusivity in query_feature_service: 'Either id (Dataset ID) or serviceUrl (Feature Service URL) must be provided; if both are given, id takes precedence.'
• Implement error handling that returns structured error objects with actionable guidance. E.g., {error: 'Invalid layer ID 5', hint: 'Call get_dataset(id) to see available layers (0-3)', retryable: false}.