Flight and Stay Search Server

Searches for one-way flights between two airports using the Duffel API, accepting departure/arrival airports, date, passenger count, and cabin class preferences. The MCP server translates user parameters into Duffel API requests, returning structured flight offers with pricing, duration, stops, and airline details. Implements parameter validation and error handling for invalid airport codes or dates.

Searches for round-trip flights by accepting outbound and return dates, applying the same multi-criteria filtering (cabin class, passenger count, airlines) to both legs. The server constructs two coordinated Duffel API requests and correlates results, returning paired flight offers with combined pricing and total journey duration. Supports flexible return dates within a date range.

Constructs multi-leg flight itineraries by chaining multiple flight searches (e.g., NYC→London, London→Paris, Paris→NYC) with user-specified dates for each leg. The server validates routing logic (arrival airport of leg N matches departure of leg N+1), executes sequential Duffel API calls, and aggregates results into a single multi-city itinerary with cumulative pricing and duration. Supports 3+ cities with flexible date ranges per leg.

Retrieves detailed information for a specific flight offer by ID, including passenger-level pricing breakdown, baggage allowances, seat availability, cancellation policies, and airline-specific terms. The server calls Duffel's offer detail endpoint and enriches results with parsed policy text and fare rules. Returns comprehensive booking-ready information without requiring a separate booking step.

Searches for accommodations (hotels, hostels, vacation rentals) using location, check-in/check-out dates, guest count, and room preferences. The server queries Duffel's accommodation API (or integrated partner like Booking.com), returning structured results with pricing, amenities, guest reviews, ratings, and availability. Supports filtering by price range, star rating, and amenity preferences (WiFi, parking, pool, etc.).

Retrieves comprehensive details for a specific accommodation by ID, including full photo gallery, guest review text and ratings, cancellation policies, house rules, and amenity descriptions. The server calls Duffel's accommodation detail endpoint and structures results for easy consumption by LLM agents or UI rendering. Includes aggregated review sentiment analysis if available.

Analyzes multiple flight offers from a search result to identify price trends, fare rules, and value propositions. The server compares base fares, taxes, fees, baggage allowances, and cancellation policies across options, returning a structured comparison with recommendations (cheapest, fastest, best value). Implements basic heuristics for value scoring (price per hour of travel, baggage flexibility, etc.).

Validates all input parameters (airport codes, dates, passenger counts, cabin classes) before executing Duffel API calls, returning structured error messages for invalid inputs. Implements checks for date logic (return date after departure, check-out after check-in), passenger count limits, and supported cabin classes. Catches and translates Duffel API errors (rate limits, invalid airports, unavailable dates) into user-friendly messages.

Defines standardized MCP tool schemas for all flight and accommodation search capabilities, enabling LLM clients to discover and invoke tools with proper type hints and parameter documentation. The server implements MCP's tool_call and tool_result protocol, managing context across multiple search requests within a conversation. Handles tool invocation errors and provides structured responses compatible with Claude, ChatGPT, and other LLM clients.