For AI Agent Developers

Travel MCP Server for AI Agents

Give your LLM agent real-world travel execution. MCP-compatible server with 9 tools for searching, booking, and managing flights. Works with Claude, GPT, and any Model Context Protocol client.

Get API Access → View Docs
MCP Tools

9 tools your AI agent can call today

Each tool follows the Model Context Protocol specification. Fully typed inputs and outputs. Your agent gets structured data it can reason about — not HTML scraping or unstructured text.

search_flights

Search flights across GDS, NDC, and LCC carriers. Returns structured offers with pricing, schedules, airlines, and fare classes. Supports one-way, round-trip, and multi-city.

book_flight

Book a flight offer by ID. Accepts passenger details and payment method. Returns PNR, e-ticket numbers, and confirmation status. Async mode for LCC bookings with webhook callback.

get_seat_map

Retrieve the seat map for a specific flight. Returns available seats with pricing, location (window/aisle/middle), and features (extra legroom, exit row).

select_seat

Assign a specific seat to a passenger on a booking. Validates availability and returns updated booking with seat assignment confirmed.

add_baggage

Add checked baggage to an existing booking. Returns available baggage options with pricing per segment and weight limits.

get_booking

Retrieve full booking details by PNR or booking ID. Includes itinerary, passenger info, ticket status, payment status, and ancillaries.

cancel_booking

Cancel a booking and initiate refund. Returns cancellation status, refund amount, and timeline. Handles airline-specific cancellation policies automatically.

search_hotels

Search hotels by destination, dates, and guest count. Returns properties with pricing, ratings, amenities, and cancellation policies. Filters by star rating, price range, and location.

get_trip_status

Real-time trip monitoring. Returns flight status, gate changes, delays, and disruption alerts. Triggers proactive notifications when the agent should intervene.

How It Works

Your agent talks. RunRelay executes.

The RunRelay MCP server exposes travel capabilities as tools that any MCP-compatible LLM can call. Your AI agent decides what to search and book based on the user's conversation; RunRelay handles the actual execution — including payment processing, ticketing, and post-booking management.

For low-cost carrier bookings (Ryanair, Wizz Air, EasyJet), RunRelay uses an async pattern: book_flight returns immediately with a pending status, and a webhook fires when the booking is confirmed. This handles the inherent latency of LCC booking systems without blocking your agent's conversation flow.

  • MCP-native — no REST wrapper needed
  • Works with Claude, GPT-4, Gemini, any MCP client
  • Typed tool schemas with validation
  • Async booking with webhook callbacks
  • Human concierge escalation for edge cases
  • Sandbox environment for testing
// MCP server configuration
{
  "mcpServers": {
    "runrelay-travel": {
      "url": "https://mcp.runrelay.io",
      "auth": {
        "type": "bearer",
        "token": "YOUR_API_KEY"
      }
    }
  }
}

// Agent calls search_flights tool
// Input:
{
  "from": "SFO",
  "to": "NRT",
  "date": "2026-05-20",
  "class": "economy",
  "passengers": 1
}

// Returns structured offers the agent
// can present and reason about
Why MCP?

The standard protocol for AI tool use.

Model Context Protocol (MCP) is the emerging standard for connecting LLMs to external tools and data sources. Instead of building custom function-calling integrations for each model provider, MCP gives you a single protocol that works across Claude, GPT, Gemini, and open-source models.

RunRelay's MCP server means your AI agent gets travel capabilities without you writing booking logic, managing supplier APIs, or handling payment flows. The agent calls a tool; RunRelay executes in the real world.

Zero Hallucination

Real data. Real bookings. No making things up.

The biggest risk with AI travel agents is hallucination — invented flight numbers, wrong prices, fake availability. RunRelay eliminates this entirely. Every tool call returns real-time data from actual airline and hotel systems. When your agent says "there's a flight at 14:30 for $450," that flight exists and that price is bookable.

  • Real-time availability from GDS/NDC/LCC systems
  • Live pricing — no cached or estimated fares
  • Actual PNR and e-ticket on booking
  • Human concierge validates edge cases
  • Structured errors — agent knows when something fails
Start Building

Give your AI agent the power to book real travel.

Get MCP server credentials, sandbox access, and documentation. Your agent can be booking flights by end of day.

Get API Access → Book a Demo