In the 2010s, IATA's New Distribution Capability (NDC) was hailed as a solution to modernize airline merchandising and distribution. Now, model context protocol (MCP) promises to simplify how artificial intelligence (AI) agents connect with external systems.
As MCP gains traction across the travel industry, stakeholders are discussing what it means for NDC—which has had a relatively sluggish uptake since its introduction.
Recently, during a BTN News Desk Session, moderator Michael Baker, executive editor of Business Travel News (BTN), asked whether MCP could enable the industry to leapfrog NDC. (BTN is a sister publication to PhocusWire within Northstar Travel Group).
NDC and MCP were designed for different reasons and operate on separate levels of the tech stack, according to Andrew Jordan, chief product and technology officer of Travelport. NDC serves as a sort of dictionary while MCP acts as a translator.
Jordan told PhocusWire that MCP has the potential to reframe NDC, driven in part by the rapid uptake of the technology among developers.
“NDC took the industry more than a decade to reach meaningful adoption,” Jordan said. “MCP has moved at a different pace: from Anthropic's announcement in late 2024 to an open standard under the Linux Foundation, with tens of thousands of active servers and backing from OpenAI, Google and Microsoft, in under 12 months.”
The problems that NDC was designed to solve are still the right problems to address in an AI-native distribution environment, he said. What’s different is how buyers and agents can access that content and how quickly integration can occur.
“The aggregation and normalization layer becomes more important, not less,” Jordan said.
During the BTN webinar titled “AI & MCP: What it means for business travel,” travel technology consultant Steve Clagg said that he sees the two acting as layers in the tech stack.
“NDC solves a content problem,” he said. “It lets airlines distribute richer, more personalized offers through intermediaries. MCP solves a connectivity problem. It lets AI agents access the system through standard interfaces.”
They aren’t conflicting from where Clagg sits. The idea that MCP could render NDC obsolete is “oversimplified,” he said.
MCP vs. NDC
The value of NDC is standardization—for airlines using it to communicate in the same dialect, Jordan said.
“NDC is an IATA-governed XML data standard,” Jordan said. “It defines a common language for how airlines describe and distribute their products, covering rich offers, ancillaries, corporate rates and real-time availability.”
Ann Cederhall, owner of LeapShift, told PhocusWire that challenges with NDC stem from varying versions of the standard.
“The problem was a myriad of versions and unbelievable slowness; about 70% of the world's airlines are on version 17.2 (second quarter of the year 2017) and servicing was made available in version 24.1 (first quarter in 2024),” Cederhall said. “I mean how can you have an API where [you] cannot service, meaning rebooking and cancelling a reservation. It is pretty useless except for booking[s] that are non-changeable.”
And it's expensive to implement a newer version or to upgrade. It requires time and resources to do so, Cederhall said.
Meanwhile, MCP is a connectivity protocol, offering a universal interface for AI agents to connect data sources and tools—including an NDC feed, without the need for a bespoke integration, Jordan continued.
MCP doesn’t eliminate the underlying complexities of distribution but can simplify how AI agents interact with external systems.
Once capabilities are exposed by an API, any compliant agent or channel can use them, Cederhall said, likening MCP to a “translation layer.”
“Where NDC defines what content looks like, MCP defines how an AI system reaches it, interrogates it and acts on it,” Jordan said.
Gee Mann, founder and CEO of Travlr ID, said during the BTN News Desk Session that because NDC standards have been deployed in different iterations across different airlines, there’s fragmentation in that world. But that’s not happening with MCP—which means LLMs can reason across separate APIs.
“I think it will drive adoption for those AI-native first booking assistants and booking tools that are trying to now use this NDC content,” Mann said.
He added that when booking tools and assistants can use NDC content across MCP servers the result could be “fantastic.”
Clagg said he does see a future in which MCP and agentic AI models could open new distribution paths, potentially reducing dependency on NDC-specific investments. But he expects corporate travel will still require authentication, negotiated rate retrieval, policy enforcement and more regardless of the delivery mechanisms used.
But Clagg cautioned travel companies against throwing away the old toy in favor of a new, shinier option: Using MCP as a reason to defer NDC implementation would be a mistake, he said.
How NDC, MCP could work together
Instead of replacing NDC, MCP could lower the barrier to entry for travel companies implementing the standard.
Clagg sees MCP and NDC working together.
“I see them as different layers in the stack,” Clagg said. “I think in most architectures that emerge, they'll coexist, and you'll have NDC aggregators, for example, that could expose their aggregated content via MCP.”
Ami Goldenberg, co-founder, CTO and board member of Oversee, said during the BTN News Desk Session that he also sees how the two could play in tandem.
“If anything, NDC and maybe even MCP should drive airlines and should drive GDS [global distribution system] as an NDC aggregator to simplify their connectivity mechanism,” he said. “We all know about NDC and the various standards, the various versions that NDC has. And this is obviously making things more complicated for others to implement.”
Goldenberg hopes that MCP could drive API providers to simplify their product.
Jordan said that as AI agents grow more capable, integration becomes more strategically significant.
One of the hurdles to NDC adoption in corporate travel has consistently been the complexity and cost of building and maintaining connections, Jordan said. But with MCP in play, that may change.
“MCP offers a faster and more flexible route to content: If an airline publishes an MCP server, an AI agent can access its inventory without conforming to that airline's specific NDC implementation,” Jordan said.