Trexin built an agentic AI proof of concept: a Voice AI Agent that creates, changes, and cancels airline reservations through natural spoken conversation, via an MCP server over a legacy GDS. The Client adopted it to Beta GA.
Our Client, a market-leading Communications Platform as a Service (CPaaS) provider, had built an advanced Voice AI Agent: real-time, natural-language conversation well beyond keypad IVR. To capture its commercial potential in travel and hospitality, they asked Trexin to build a proof of concept integrating that agent with a large, legacy Global Distribution System (GDS): creating, modifying, and canceling airline reservations entirely through unscripted, spoken conversation.
We built on the Model Context Protocol (MCP), the open standard that lets LLM-based systems connect to external tools and data consistently and at scale. The Client’s Voice AI Agent acted as the MCP Host; Trexin developed a new MCP Server in Python that wrapped the GDS’s legacy REST APIs and ran in a Docker container in the Client’s private cloud. The Host discovered the booking tools the server exposed, enriched the model’s context, and (from a caller’s natural language) selected and executed the right tools to fulfill each request.
A working agentic capability that handled end-to-end reservation workflows (airport-code resolution, itinerary lookup, booking creation, and changes) all driven through natural-language voice. Given its quality and stability, the Client adopted the POC as a Beta release to General Availability for their own customers.
We turned an advanced capability into a working, adoptable system: agentic AI doing real work, not a parlor trick.
case study
case study
event
Tell us what you're trying to do. A senior practitioner will read it.
Talk to us