Server has 8 tools with complete input schemas and descriptions in German. All tools are READ_ONLY and follow verb_noun naming (lookup_, search_, get_, check_, find_). Descriptions are present (avg ~150 chars) and explain WHAT the tool does. However, parameter descriptions lack detail on constraints (e.g., barcode format, max_results bounds not enforced in schema). Output schemas are not formally documented, responses are inferred from code. Error handling returns generic error dicts without recovery guidance. No tool annotations (readOnlyHint, etc.). Composition is sound: tools are single-purpose and chainable via barcode. Missing: enum constraints for categorical params, explicit output schema documentation, actionable error messages.

• Add formal JSON Schema output definitions for each tool. Document all response fields, types, and whether they are always present or conditional.
• Add minItems/maxItems and pattern constraints to schema. E.g., max_results: {type: integer, minimum: 1, maximum: 20}; barcode: {type: string, pattern: '^\d{13}$'}.
• Enhance error responses with actionable recovery hints. Instead of {fehler: 'Produkt nicht gefunden'}, return {error: 'Product not found', suggestion: 'Try search_food(query=...) to find the product by name', retryable: false}.
• Add readOnlyHint: true to all tool schemas to signal to LLMs that these tools are safe to retry and have no side effects.
• Translate parameter descriptions to English or add English descriptions alongside German. Specify format constraints explicitly: 'barcode: 13-digit EAN code (e.g., 3017620422003)' instead of relying on examples.