Single tool 'search_listings' has a well-structured schema with proper type definitions, clear descriptions, and appropriate constraints (limit 1-100, offset ≥0). Tool name follows verb_noun pattern. Description is concise (118 chars, within 10-1024 baseline). Input schema uses Pydantic with validation. However, output schema is not documented in the visible code, only structuredContent is mentioned without field definitions. Tool annotations are present (readOnlyHint=true). Error handling and recovery guidance are not visible in the provided code.

• Document the output schema explicitly. Define the structure of each listing object (e.g., id, title, price, location, lat, lng, url) and return type of the payload dict. Include this in the tool's docstring or as a separate schema definition.
• Add error handling guidance to the tool description: 'Returns empty results if no listings match the query. If the search service is unavailable, returns an error with retry guidance.'
• Constrain the 'query' parameter: add min/max length (e.g., 3-500 chars), and clarify that it accepts natural language (e.g., '2-bedroom apartment in Zurich near public transport').
• Explain pagination in the tool description: 'Use limit and offset to paginate results. If offset >= total count, returns empty results. Call with offset=0 to start, then increment by limit to fetch subsequent pages.'
• Add a 'total_count' field to the response payload so agents know how many results exist and can decide whether to paginate or refine the query.