Three tools with complete input schemas and descriptions. Naming follows verb_noun convention (search_, dataset_, query). Descriptions are adequate (100-200 chars) but lack dependency hints and prerequisites. Parameters have types and descriptions, but lack enums for constrained inputs (e.g., 'limit' should enforce 1-100 via schema, not just description). Output schemas are undocumented, LLMs cannot predict response structure for chaining. Error handling is minimal (generic HTTP errors only). No security annotations or permission gates. Composition is sound: search → info → query workflow is logical.

• Document output schemas for each tool. Example for search_datasets: 'Returns {datasets: [{id, title, theme, record_count}], total_count, limit, offset}'. This enables LLMs to extract dataset_id for chaining.
• Add JSON Schema constraints: limit with minimum: 1, maximum: 100; offset with minimum: 0. Remove range descriptions from text, let the schema enforce it.
• Enhance error messages with recovery guidance. E.g., catch invalid dataset_id and return: 'Dataset not found. Call search_datasets(query="<keyword>") to find valid dataset IDs.'
• Add dependency hints to descriptions. E.g., dataset_info: 'Call this after search_datasets to learn column names before querying.' query: 'Requires dataset_id from search_datasets and optional column names from dataset_info.'
• Add tool annotations to inputSchema: readOnlyHint: true and idempotentHint: true for all three tools. This signals to agents that these calls are safe to retry and do not modify state.