The Dubai-based aggregator has layered a Model Context Protocol interface onto its Iris platform, giving AI systems machine-readable access to more than 60 airlines through a single protocol. The technical move is sound. The harder question is whether the industry is ready to use it.
There is a version of this announcement that reads as straightforward product marketing: an aggregator adds an AI-friendly protocol to its platform, issues a press release, and positions itself ahead of a technology wave. TPConnects has done exactly that. But the underlying architecture deserves a closer read, because what the company is actually describing is a meaningful shift in how distribution infrastructure gets consumed, not just by travel sellers, but by the automated systems that are increasingly making decisions on their behalf.
Iris already normalises content across more than 60 NDC-enabled carriers, LCC inventory, and all four major GDS platforms into a single interface. That aggregation layer has been the core product for several years. What changed last week is that TPConnects added Model Context Protocol on top of that existing infrastructure, a standardised, machine-readable layer that allows AI agents, automated booking tools, and conversational commerce systems to interact with airline content without requiring custom integrations for each carrier.
The technical logic is straightforward. Every aggregator in the market solves the same problem at the content level: take fragmented, inconsistent airline data from multiple channels and present it in a normalised format. What very few have done is expose that normalised content through a protocol that AI systems can discover and interrogate dynamically. The distinction matters because the next generation of travel tools, autonomous booking agents, AI-powered TMC platforms, conversational retail interfaces, do not interact with APIs the way a developer does. They need to understand what capabilities exist, what data structures look like, and what actions are available, without someone hardcoding that knowledge in advance. MCP is designed for exactly that pattern.
Praveen Kumar, Co-founder and CTO at TPConnects, described the architectural intent clearly: when a carrier is added to Iris or an existing airline introduces new ancillaries or servicing rules, MCP-reading systems detect those changes automatically. No configuration update, no deployment cycle, no manual intervention. The system learns what is available and makes it accessible immediately.
That self-discovery property is the genuinely interesting part of the announcement, and it is also where the real-world complexity begins.
NDC implementations are notoriously inconsistent across carriers. An airline that is technically NDC-certified may support offer creation but not order management. Another may handle exchanges but not refunds through the NDC channel. A third may have ancillary content available in shopping but not in servicing. Aggregators like TPConnects deal with this by building carrier-specific logic into their normalisation layer, essentially absorbing the inconsistency so that the seller-facing interface behaves predictably. The question the MCP announcement raises is how that carrier-specific complexity gets represented in the protocol. If MCP exposes a clean, consistent interface but the underlying servicing behavior varies significantly by carrier, the self-learning capability discovery that TPConnects is promoting has real limits.
This is not a criticism unique to TPConnects. It is the structural problem that every aggregator faces when making normalisation claims. The value of the platform depends entirely on how well the normalisation holds under the edge cases, the disruption scenarios, the ancillary modifications, the mid-booking carrier changes that are uncommon enough to escape testing but common enough to create operational pain at scale.
What TPConnects has done is make a credible architectural bet. MCP is gaining real traction in enterprise software, and travel distribution is a natural fit for the protocol given the volume and structure of the data involved. The company is not the only one moving in this direction. The broader race to become the AI-ready distribution layer is well underway across the aggregator, GDS, and airline technology markets. But TPConnects has moved from concept to production deployment, which puts it ahead of most of the field on this specific capability.
For TMCs and OTAs evaluating the Iris MCP integration, the relevant questions are operational rather than architectural. Does the protocol cover the full transaction lifecycle including post-booking servicing, or is it currently scoped to shopping and booking? How does the system behave when airlines push breaking changes to their NDC implementations? What happens to connected AI agents in the gap between a carrier update and the MCP layer reflecting it?
TPConnects has not published detailed answers to those questions, which is typical for an early-stage production release. The answers will emerge from how the product performs at scale over the next twelve months.
The Iris MCP integration is available to customers now. TPConnects can be reached at tpconnects.com.
Travel Distribution News covers airline distribution, NDC adoption, GDS dynamics, travel payments, and emerging markets.
