Three tools with clear verb-noun naming (search_datasets, dataset, query_resource) and reasonable descriptions (120-180 chars). All have input schemas with type definitions and required fields properly marked. However, parameter descriptions lack actionable constraints (no enums, ranges, or format specs), output schemas are undocumented, and error handling is minimal. No tool annotations (readOnlyHint, etc.). Descriptions are adequate but generic, they state WHAT but not WHEN or WHY to use each tool relative to others.

• Document output schemas for all three tools. Specify that search_datasets returns {results: [{id, name, title, organization, resources: [{resource_id, ...}]}], total_count, ...} and dataset returns {id, name, title, resources: [...]}. This enables agents to chain calls correctly.
• Add enum or pattern constraints to parameters. Change 'query' description to 'Keyword(s) to search; pass "*" or empty string to list all datasets.' Add minItems/maxItems to limit parameter: 'Max datasets to return (1-100, default 20).'
• Add tool annotations to the schema. Set readOnlyHint: true for all three tools since they only read data and have no side effects. This signals to agents that these calls are safe to retry.
• Enhance error messages with recovery guidance. When dataset lookup fails, catch the error and return: 'Dataset not found. Try search_datasets(query="...") to find the correct dataset ID.' When a resource has no DataStore, return: 'This resource does not support DataStore queries. Check the dataset metadata for queryable resources.'
• Add dependency hints to descriptions. Update dataset description to: 'Show full metadata for a Finland Open Data dataset by id or name. Tip: use search_datasets() first to find a dataset ID if you only have a keyword.' This guides multi-step planning.