Case Study December 8, 2025
Bringing Agentic AI To Life For Airline Reservations
Trexin helped an innovative telecommunications leader integrate its Voice AI Agent capabilities with a travel & hospitality booking system.
Business Driver
Our Client, a market-leading Communications Platform as a Service (CPaaS) provider, developed an advanced Voice AI Agent capability that allowed its global communications and connectivity customers to deploy leading-edge virtual assistants that were capable of having real-time conversations, understanding what customers say using their own language and phrasing, and responding appropriately. These capabilities far surpassed traditional interactive voice response (IVR) systems that were limited to keypad inputs and pre-defined menu options. Eager to capture the full commercial potential of this innovation in the travel and hospitality sector, our Client asked Trexin to collaborate on a proof of concept (POC) to integrate their Voice AI Agent with a large, legacy Global Distribution System (GDS) to demonstrate the creation, modification, and cancelation of airline reservations entirely through unscripted, spoken conversation.
Approach
Our implementation approach focused on the application of the Model Context Protocol (MCP), an open standard that allows AI systems with large language models to connect and interact with external data sources, tools, and services in a universal, consistent, and scalable manner. Our Client’s Voice AI Agent, which served as the MCP “Host”, incorporated an MCP Client to manage the connections and communications between the Host and MCP Servers. In this POC, Trexin developed a new MCP Server in Python code that wrapped the legacy REST-based APIs of the GDS and deployed it within a Docker container running in our Client’s private cloud environment supporting containerized applications. Per the MCP specifications, the Host (via the MCP Client) established a connection to the MCP Server, which was designed to allow the Host to discover the available tools and resources related to booking, changing, and canceling an airline reservation. The Host then enriched the AI model’s context with this information, and based on a user’s spoken natural language, the AI model intelligently selected and executed the specific tools required to fulfill users’ requests.
Results
Trexin was able to successfully build an Agentic AI capability to create, modify, and cancel airline reservations through conversational, spoken language. The POC integrated our Client’s Voice AI Agent with a new MCP Server, enabling real-time tool calls into a leading global airline booking and scheduling platform used across the travel industry. The solution reliably demonstrated end-to-end reservation workflows including airport code resolution, itinerary lookup, booking creation, and itinerary changes, all driven through natural language voice interaction. Given the quality and stability of this POC, it was effectively adopted by our Client as a Beta release to General Availability for their customers’ use.
