Most travel sellers think the hard part of airline content aggregation is the volume. Managing two NDC versions for a single airline, multiplied across 25-plus carriers, each with different schemas, different feature sets, and different seasonal variations, that’s where the real complexity lives. Aggregation is a solved problem. Normalization at that scale isn’t.
In the latest instalment of the TPConnects Interview Series, Praveen Kumar, CTO and Co-Founder of TPConnects, breaks down what MCP actually means for travel distribution, why the intelligence is in the routing rather than the aggregation, and what changes when an AI agent reads available functions directly instead of navigating a traditional API integration. The conversation covers the Iris MCP server now live across NDC, LCC, and GDS sources, what zero-code integration looks like in practice, and how search data, historically discarded, becomes a commercial asset.
About Praveen Kumar
Praveen Kumar is the CTO and Co-Founder of TPConnects, where he leads the engineering and product architecture behind Iris, the travel seller aggregation platform, and Astra, the airline retailing layer. He is the architect behind TPConnects’ MCP layer, a project that began as a proof of concept at AWS Summit and has since become the foundation of the company’s AI distribution strategy.
What the interview covers
The content challenge that goes beyond aggregation
Managing two NDC versions for a single airline, multiplied across 25-plus carriers with different schemas, different feature sets, and different seasonal variations. Praveen explains why normalization is where the complexity actually lives, and what that means for any OTA trying to keep pace with airline content changes without a permanent integration backlog.
What a unified MCP server looks like across NDC, LCC, and GDS
Iris consumes NDC, LCC, and GDS sources and surfaces them through a single aggregated MCP server. Praveen walks through how the system routes queries to the right airline source based on context, what that means for search performance, and how it changes the integration overhead for travel sellers connecting to new content.
Zero-code integration and what it changes
An MCP client takes an endpoint, adds credentials, and the connection is live. No integration project, no development sprint. Praveen explains what changes when the AI reads available functions directly, and why a new airline can go live in a week rather than three months. Iris has added over ten airlines in a few months using this approach.
Search data as a commercial asset
Every anonymous search is wasted context in the old model. Praveen explains how storing search patterns by route, season, and traveler type builds a knowledge base that anticipates demand, and how that overhead becomes commercial intelligence rather than sunk cost.
The interview is available on demand. If you’re a travel agency owner, OTA product lead, TMC operations team, or consolidator managing multiple content integrations today, or evaluating a move to a normalized aggregation layer, this is the conversation to watch.
