A multi-agent travel planning system that integrates flight search, hotel recommendations, weather forecasts, and itinerary generation using LangGraph and MCP tools
Server has 5 tools with minimal definition quality. All tools have basic descriptions (10-40 chars), but lack parameter descriptions entirely. Input schemas are present but sparse: get_current_weather and get_forecast have single string params with descriptions; tavily_search has a query param with description; list_airports and list_airlines have empty input objects with no parameters. No output schemas documented. No error handling guidance. Tool names follow verb_noun convention (get_*, list_*, search_*) which is correct, but descriptions are too brief to guide LLM selection. No tool annotations (readOnlyHint, etc.). Composition is reasonable (each tool does one thing), but parameter documentation is critically absent.
Get current weather for a city
Get weather forecast for a city
List available airlines from aviation data
List available airports from aviation data
Search the web using Tavily search API
list_airports and list_airlines have empty input schemas (no parameters) with no documentation of what they return, pagination support, or result limits. LLMs cannot plan downstream calls without knowing output structure.
All tool descriptions are under 40 characters and lack context on WHEN to use each tool vs alternatives. 'Get current weather for a city' does not explain when to call this vs get_forecast, or what format the response uses.
No output schemas documented for any tool. LLMs cannot infer what fields to expect (e.g., does get_current_weather return temperature, humidity, wind_speed? What are the field names and types?). This forces agents to guess or fail.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | 2026-07-28+ | v2 |
No error handling guidance. Tools do not document what happens on invalid input (e.g., non-existent city), network failures, or rate limits. Agents have no recovery path.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All tools are read-only but this is not declared, forcing LLMs to infer safety properties.