MCP server for travel aggregation - flights, hotels, and more
This server has moderate structural quality with good naming and documented schemas, but critical gaps in parameter descriptions, output schema documentation, and error handling guidance. Seven tools are explicitly registered with consistent verb-noun naming (search_*, track_*, get_*) and non-trivial descriptions (100-150 chars). However, nearly all tools have underdescribed or missing parameter constraints, no documented output schemas, and no error recovery guidance. The schema definitions are present and reasonably detailed for input parameters, but lack critical information about return structure, pagination, required fields in responses, and how to handle edge cases. Most tools are READ_ONLY (safe), reducing risk, but the lack of error categorization and recovery hints would force an LLM to guess on failures.
Find the cheapest dates to fly between two airports. Useful for flexible travel planning.
Get detailed information about an airport including location, timezone, and facilities.
Get all flights on a specific route. Shows all airlines flying between two airports.
Search for airports by city name or location. Returns airport codes, names, and locations.
Search for flight offers between two airports. Returns flight options with prices, airlines, and schedules.
Search for hotels in a city with availability and pricing. Returns hotel options with rates, amenities, and location.
No output schema documentation for any tool. LLMs cannot plan downstream operations or extract required fields without knowing the response structure. Example: search_flights returns flight data, but critical fields like 'price', 'airline', 'departure_time', 'flight_id' are not documented.
Missing or incomplete parameter descriptions prevent LLMs from understanding constraints and expected formats. Example: 'track_flight' has two parameters (flight_iata, flight_icao) with no indication that at least one is REQUIRED, no format specification, and no guidance on which to prefer when both are available.
No error handling or recovery guidance. Tools lack actionable error messages. If search_flights returns a 400 (invalid airport code), the LLM has no guidance to call search_airports first. No categorization of errors as retryable vs user-fixable.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
Track a flight in real-time. Get current status, location, altitude, and estimated arrival.
No pagination or result limit documentation. Tools like search_flights and search_hotels have max_results parameters, but no guidance on default limits, maximum safe values, or how to handle pagination tokens/cursors if results exceed limits.
Parameter 'return_date' in search_flights is marked optional ('optional for one-way flights') but lacks formal conditional logic. No schema constraint prevents passing both departure and return for a one-way flight, leaving validation to backend.
Track_flight parameter 'flight_iata' vs 'flight_icao' creates ambiguity. Neither is marked required; description does not clarify whether both, one, or at least one is needed. LLM must guess which to provide.
No tool idempotence declaration. Tools like track_flight and search_flights should document whether repeated calls with identical input always return the same result (safe to retry) or if state changes (date-dependent pricing, flight status updates). Lack of clarity forces LLMs to treat retries as risky.