DEV Community

iamTheDev
iamTheDev

Posted on

Embedding MCP Into Hotel CRS: Why This Is Hotel Distribution's "Disintermediation"

Anyone in hotels knows the story. For years, distribution power has sat with OTAs. Hotels sell rooms through Ctrip, Fliggy, and similar channels, and each order loses a big commission. Hotels want direct sales but can't make it work—users are conditioned to compare and book on OTAs. The hotel's own website gets no traffic.

But recently I noticed something: some hotel tech companies are embedding MCP directly into hotel CRS (Central Reservation System). I think this could shake OTA's position in hotel distribution. Let's talk about what this means.

First: What Is a CRS? Why Does Embedding MCP Matter?

CRS is a hotel's Central Reservation System—core IT that holds all room status, pricing, and orders.

Before, hotels wanted to sell rooms on OTAs. The hotel CRS syncs inventory and pricing to OTAs through various integration methods. A user books on the OTA, and the OTA passes the order back to the hotel CRS. That middle layer is distribution.

What does "embedding MCP into CRS" mean? Building an MCP interface directly into the CRS. The hotel CRS exposes an MCP server. Any MCP-compatible AI agent can call the CRS directly—query inventory, check pricing, place orders.

What does this mean? Hotels can sell rooms directly to AI agents without going through an OTA. User tells an AI assistant to book a hotel. The AI calls the hotel's MCP directly and places the order. No OTA middle layer.

Traditional Distribution: Layered Markups, Hotels With No Leverage

Review the traditional hotel distribution model:

Hotel CRS
  ↓ (sync inventory and pricing)
Wholesaler / distributor
  ↓ (re-sync)
OTA platform (Ctrip / Fliggy / Booking)
  ↓ (traffic distribution)
User books
Enter fullscreen mode Exit fullscreen mode

Every layer takes a cut. Wholesalers earn one layer. OTAs earn another. The hotel ends up with 60-70% of what the user paid.

Worse: OTAs hold user traffic. Hotels want to sell rooms, they beg OTAs. OTA decides your ranking, OTA decides your commission rate. Hotels have no leverage.

Hotels didn't try direct sales. Official website, app, mini-program. The results were weak—because users book hotels by opening an OTA first. The hotel's own site doesn't come to mind.

The New Model: Hotels Connect MCP, AI Agents Book Direct

After embedding MCP into CRS, the flow looks like this:

Hotel CRS (built-in MCP server)
  ↓ (standard MCP protocol)
AI assistant / travel agent (various C-end products)
  ↓
User books directly
Enter fullscreen mode Exit fullscreen mode

See the difference? The wholesaler and OTA layers are gone. Hotel connects to AI agent directly. AI agent connects to user directly.

That's why it's "disintermediation"—MCP removes the middle distribution layer.

And this model has an advantage: users don't download the hotel's app or remember its website. They tell their usual AI assistant "book me a hotel near the Bund in Shanghai." The AI calls the hotel's MCP directly and books it.

For users, the experience is about the same as using an OTA. Maybe even easier—no comparing, no scrolling. AI handles it. For hotels, they save a big commission.

RollingGo's hotel MCP already aggregates significant hotel inventory. Behind it: 110K+ direct-contracted hotels and 2M+ global properties. Anchored upstream of OTA supply, with strong pricing on international hotels plus Shanghai, Hong Kong, Japan, Korea, and other popular destinations. Direct inventory with real-time price confirmation. What you see is what you can book. Completely free, no call-volume limits. Supports 40+ major LLM clients. Provides an all-in-one booking skill for ClawHub and Coze. Developer partners set country-based markup rates and earn commission.

Over 2,000 agents already connect: regional tourism platforms, AI earbuds, trip-planning apps. This direct-connect model will keep growing. Want to try the disintermediated experience? Follow the quick-start guide and get a key.

Setup in Codex:

[mcp_servers.RollingGo-Hotel]
url = "https://mcp.rollinggo.ai/mcp"
http_headers = { "Authorization" = "Bearer YOUR_API_KEY" }
Enter fullscreen mode Exit fullscreen mode

Restart Codex and you're calling hotel search, room detail, and price confirmation directly. Hotel distribution disintermediation starts with MCP direct connect.

Why Does This Shake OTA Position?

Why do OTAs hold power in hotel distribution? Two things: traffic entry and transaction closure.

Both are being challenged by AI agents:

First, the traffic entry is changing. Before, users opened an OTA app to book hotels. What about the future? Users open an AI assistant and say one sentence. The entry point shifts from "OTA app" to "AI assistant." OTA traffic advantage weakens.

Second, transaction closure is changing. Before, users booked on OTAs, and OTAs held payment, orders, and user data. What about the future? Users book through AI assistants, and AI agents hold the transaction closure. OTAs get bypassed again.

This doesn't mean OTAs die tomorrow. Years of supply chain investment and user trust don't disappear instantly. But the trend is: a new direct path is emerging between hotels and users. That path is MCP.

If this path holds, OTA's middle position gets increasingly awkward. It goes from "must-pass" to "one optional channel."

What Does This Mean for Hotels?

This is good news for hotels.

Before, hotels wanted direct sales but had to build traffic, apps, and user operations themselves. High cost, weak results. Now with MCP, hotels connect their CRS to MCP and they appear in every AI assistant. Users book through AI and find your hotel.

MCP opens a new direct channel for hotels. You don't build traffic. AI agents naturally recommend your hotel to users.

And this channel has better margins—no OTA commission to pay. Same room sold, more money stays with the hotel.

This does put demands on hotels: your CRS needs MCP support, and your inventory and pricing need to be accurate. If AI recommends your hotel and users click through to find no rooms or wrong prices, trust breaks.

How Will OTAs Respond?

OTAs aren't stupid. They won't sit still. I see a few likely responses:

One, build MCP themselves. If MCP enables disintermediation, I'll build an MCP server too and expose my inventory as MCP. AI agents can connect to my OTA's MCP. I'm still in the middle.

Two, build their own AI assistant. OTAs are building AI travel assistants. They want users to stay in their ecosystem. User books hotels through my AI assistant, using my inventory, and I still earn commission.

Three, partner with LLM providers. Work with GPT and Claude to prioritize OTA inventory recommendations. Even through AI assistants, OTAs stay in the mix.

OTAs will defend their position. But regardless of how they defend it, the distribution chain is shifting because of MCP.

Embedding MCP into hotel CRS looks small but means a lot. It could restructure hotel distribution. Before, hotels and users had to go through OTA intermediaries. Now with MCP, hotels and users can connect directly through AI agents.

That's disintermediation. It won't happen overnight, and OTAs won't surrender without a fight. But the trend is there. Hotel distribution power might be slowly shifting from OTAs back to hotels themselves.

Do you think hotel direct sales can beat OTAs? Drop a comment.

Top comments (0)