Five tools with complete input schemas and clear descriptions. All tools follow verb_noun naming (search_*, list_*, get_*). Descriptions are well-written (100-200 chars), explaining what each tool does and when to use it. Parameters have types and descriptions. However, output schemas are not documented, LLMs cannot see what fields to expect from responses. No error handling guidance. No enum constraints on 'lang' parameter despite accepting only 'en'/'pl'. Missing idempotency hints and pagination guidance in descriptions.

• Document output schemas for all tools. Example: 'Returns {data: [{id, type, attributes: {title, notes, ...}, relationships: {...}}], links: {self, next, last}, meta: {...}}'. Include field types and meanings.
• Convert 'lang' parameter to an enum: {type: 'string', enum: ['en', 'pl'], description: 'Language hint...'}. Prevents invalid values.
• Add error handling guidance to tool descriptions. Example: 'If dataset not found, try search_datasets() with a partial title. If resource_id is invalid, call list_resources() first to find valid ids.'
• Clarify pagination in descriptions: 'Returns paginated results (default 20 per page). Check the 'links.next' field in the response to fetch the next page. If 'links.next' is absent, you have reached the end.'
• Explain JSON:API response structure in the server's top-level documentation or in each tool description: 'Responses use JSON:API format. Extract data from the 'data' array, then access 'attributes' for field values and 'relationships' for linked resources.'