Four tools with complete schemas and descriptions. Naming follows verb_noun pattern (list_, get_, search_). Descriptions are concise (10-50 chars) and action-oriented. All parameters have types and descriptions. Output schemas are documented via Pydantic models (Weather, ForecastDay, Flight). Input validation is present (city lookup, range checks). However, descriptions are at the lower end of the 10-1024 char baseline (avg 194 chars for A+ tools); some lack context on WHEN to use vs similar tools. Error messages are generic ValueError rather than recovery-guided. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all being read-only. Composition is clean, each tool has one responsibility.

• Expand tool descriptions to 50-150 chars. Add WHEN-to-use guidance: 'Use get_weather for today's conditions; use get_forecast for multi-day planning. Only supported destinations work, call list_destinations first if unsure.'
• Enhance error messages with recovery guidance. Instead of 'Unknown city', return 'Unknown city "Paris". Supported cities: amsterdam, barcelona, ... Try list_destinations() to see all options.'
• Add tool annotations via MCPServer.tool() decorator or response metadata. Mark all four tools with readOnlyHint=true to signal they are safe for speculative execution.
• Replace example values in descriptions with formal constraints. E.g. 'City name (string, case-insensitive, must be in supported list)' instead of 'e.g. Amsterdam'.
• Document parameter dependencies. E.g. 'days must be 1-7; default 3. units applies to both high_c and low_c in response.'