Three tools with complete schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. All tools have action verbs (search_, dataset_, query) and non-empty descriptions (avg 156 chars). However, 7 of 11 parameters lack type constraints (enums, ranges), and no output schemas are documented. Error handling is minimal, raw HTTP errors with no recovery guidance. The 'query' tool accepts free-form ODSQL strings without validation hints, inviting malformed queries. Composition is sound: search→info→query workflow is logical and IDs chain correctly.

• Document output schemas for all three tools. Specify: search_datasets returns {datasets: [{dataset_id, title, theme, record_count}], total_count}; dataset_info returns {fields: [{name, type, description}], theme, record_count}; query returns {records: [...], total_count}.
• Add ODSQL syntax hints to where/select/group_by/order_by descriptions. E.g., 'ODSQL WHERE clause (e.g., year >= 2020 AND city = "Lyon"). See OpenDataSoft docs for syntax.'
• Enhance error messages with recovery guidance. E.g., 'Dataset not found. Try search_datasets() to find available datasets.' or 'Invalid ODSQL: check syntax at [docs-link].'
• Add enum or pattern constraints for limit parameter. Explicitly state: 'Limit must be 1 - 100 (default 20). Values outside this range are clamped.'
• Return pagination metadata (total_count, has_more) in search_datasets and query responses so LLM can determine if additional calls are needed.