News • Jun 1, 2026

The MCP Server That Hospitality’s AI Future Runs On

avatar Barbara Domingues
Author: Eric Sniff, CTO and Chad Kimner, COO, Visiting Media

 

Five weeks ago, a hotel GM asked us a question that we’ve been hearing more and more of as of late: “How can you help us be sure AI agents recommend our property?”Then it happened again. Then it happened a third time. Three unsolicited asks in few weeks is a pattern. Not a market category yet. Not a finished roadmap. A pattern.

The question underneath it is simple. If a traveler asks an AI assistant where to stay, where does the assistant get the answer? It can read public web pages. It can summarize reviews. It can infer from whatever structured data happens to be available. What it usually can’t do is query the hotel’s richest content: the immersive media, property details, room-level context, venue assets, and experience data the hotel already maintains.

The internal version is just as simple. When your own team asks their AI assistant about your properties, where does it get the answer? It can write the email. It can summarize the report. It can pull from whatever’s already in their drive. What it can’t do is reach your live Visiting Media platform data: can you add a tour to an email, who uploaded what last month, how tours are performing this week. The MCP closes that gap, scoped to what each person can already see: the whole portfolio for above-property users, a single hotel for everyone else.

MCP stands for Model Context Protocol. Anthropic created it. The plain-English version: MCP gives an AI agent a standard way to query a system and get a structured answer back. Instead of scraping a page or hallucinating from stale text, the agent can call a real interface with real permissions.

For hospitality, that matters because the product is physical. A hotel isn’t a SKU in a clean database. It’s rooms, views, meeting spaces, renovations, amenities, packages, seasons, brand standards, ownership constraints, and sales context. Most of the detail that matters to a buyer is either trapped in media libraries or flattened into generic copy.

Michael Chamerski on our Engineering team (and his fleet of agents) team shipped our working VMP API MCP server with OAuth 2.1 in roughly 10 days.

That sentence has two parts.

The engineering part is the server: authenticated, scoped, built against a real API. OAuth 2.1 matters because access has to be permissioned. An agent should only see what the customer has allowed it to see. If a key is scoped to one property group, that’s all the agent sees. Nothing expands at the agent layer.

The business part is the timing. Carolann Neale in our product marketing team started sharing setup documentation with the CX team last week, so customer onboarding can begin. That moves this from a technical demo into something customers actually use.

v1 is deliberately narrow. Read, learn, trust. Write actions aren’t the right first move. Before an agent updates property content, changes tags, or modifies customer-owned assets, it has to earn trust on read access first. Can it answer correctly? Can it say when it doesn’t know? Can customers understand what it saw and why it answered the way it did?

That’s the spec for the first version. The partner pattern is the strategic part. “Brand-level MCPs become the distribution surface. Visiting Media becomes the content layer underneath.” – Chad Kimner

Chad’s framing here is useful, because this is a GTM workflow story, not just an integration. Hotels have spent years producing better content, then pushing it into channels that were never built to preserve the context. The handoff is lumpy. The asset might be rich, but the downstream system sees a thumbnail, a caption, or a generic amenity field.

MCP changes the handoff. It lets the content answer questions. That doesn’t mean every AI assistant on the internet suddenly knows which suite has the best terrace or which meeting room works for a 70-person reception. There are still gaps. Brand MCPs are early. Customer onboarding is early. The standards for agent discovery, ranking, and attribution are still forming. We should be honest about that.

But the direction is clear enough to build against. Hotels want to be found by agents because their buyers are starting to use agents. Partners want a standard surface because every one-off integration gets expensive to maintain. Brands want distribution control because the last generation of digital travel taught everyone what happens when the demand layer is owned somewhere else.

We didn’t build MCP. We build the part underneath it that’s actually hard in hospitality: the immersive content, the property context, the workflows that make all of it usable to a buyer.

The hotel industry has spent 20 years optimizing for search engines. The next 10 will be about optimizing for AI agents.

Contact our team to learn more about what this unlocks for you today.

News • Jun 1, 2026

The MCP Server That Hospitality’s AI Future Runs On

avatar Barbara Domingues
Author: Eric Sniff, CTO and Chad Kimner, COO, Visiting Media

 

Five weeks ago, a hotel GM asked us a question that we’ve been hearing more and more of as of late: “How can you help us be sure AI agents recommend our property?”Then it happened again. Then it happened a third time. Three unsolicited asks in few weeks is a pattern. Not a market category yet. Not a finished roadmap. A pattern.

The question underneath it is simple. If a traveler asks an AI assistant where to stay, where does the assistant get the answer? It can read public web pages. It can summarize reviews. It can infer from whatever structured data happens to be available. What it usually can’t do is query the hotel’s richest content: the immersive media, property details, room-level context, venue assets, and experience data the hotel already maintains.

The internal version is just as simple. When your own team asks their AI assistant about your properties, where does it get the answer? It can write the email. It can summarize the report. It can pull from whatever’s already in their drive. What it can’t do is reach your live Visiting Media platform data: can you add a tour to an email, who uploaded what last month, how tours are performing this week. The MCP closes that gap, scoped to what each person can already see: the whole portfolio for above-property users, a single hotel for everyone else.

MCP stands for Model Context Protocol. Anthropic created it. The plain-English version: MCP gives an AI agent a standard way to query a system and get a structured answer back. Instead of scraping a page or hallucinating from stale text, the agent can call a real interface with real permissions.

For hospitality, that matters because the product is physical. A hotel isn’t a SKU in a clean database. It’s rooms, views, meeting spaces, renovations, amenities, packages, seasons, brand standards, ownership constraints, and sales context. Most of the detail that matters to a buyer is either trapped in media libraries or flattened into generic copy.

Michael Chamerski on our Engineering team (and his fleet of agents) team shipped our working VMP API MCP server with OAuth 2.1 in roughly 10 days.

That sentence has two parts.

The engineering part is the server: authenticated, scoped, built against a real API. OAuth 2.1 matters because access has to be permissioned. An agent should only see what the customer has allowed it to see. If a key is scoped to one property group, that’s all the agent sees. Nothing expands at the agent layer.

The business part is the timing. Carolann Neale in our product marketing team started sharing setup documentation with the CX team last week, so customer onboarding can begin. That moves this from a technical demo into something customers actually use.

v1 is deliberately narrow. Read, learn, trust. Write actions aren’t the right first move. Before an agent updates property content, changes tags, or modifies customer-owned assets, it has to earn trust on read access first. Can it answer correctly? Can it say when it doesn’t know? Can customers understand what it saw and why it answered the way it did?

That’s the spec for the first version. The partner pattern is the strategic part. “Brand-level MCPs become the distribution surface. Visiting Media becomes the content layer underneath.” – Chad Kimner

Chad’s framing here is useful, because this is a GTM workflow story, not just an integration. Hotels have spent years producing better content, then pushing it into channels that were never built to preserve the context. The handoff is lumpy. The asset might be rich, but the downstream system sees a thumbnail, a caption, or a generic amenity field.

MCP changes the handoff. It lets the content answer questions. That doesn’t mean every AI assistant on the internet suddenly knows which suite has the best terrace or which meeting room works for a 70-person reception. There are still gaps. Brand MCPs are early. Customer onboarding is early. The standards for agent discovery, ranking, and attribution are still forming. We should be honest about that.

But the direction is clear enough to build against. Hotels want to be found by agents because their buyers are starting to use agents. Partners want a standard surface because every one-off integration gets expensive to maintain. Brands want distribution control because the last generation of digital travel taught everyone what happens when the demand layer is owned somewhere else.

We didn’t build MCP. We build the part underneath it that’s actually hard in hospitality: the immersive content, the property context, the workflows that make all of it usable to a buyer.

The hotel industry has spent 20 years optimizing for search engines. The next 10 will be about optimizing for AI agents.

Contact our team to learn more about what this unlocks for you today.