Mews and AI: Automating Hotel Operations with an Agent
By Alexandre Saint-Jean

Audio version
Audio version produced by text-to-speech from the article. Our AI charter
A hotel PMS such as Mews centralises bookings, pricing, check-in and payment. Two of its building blocks already have AI built in natively: pricing and check-in. Everything else runs through its open API. This article is part of our overview of AI in hospitality and restaurants.
What does AI already do in Mews with nothing extra connected?
Mews ships AI natively on two specific functions: automatic pricing through Autopilot, and self-service check-in through Kiosk. These are modules of the PMS, not add-ons, available with no extra integration for hotels already running the platform.
Everything else (follow-ups, cross-property reporting, review responses) is not part of Mews's native offering and requires going through the PMS's API.
How does automatic pricing work with Mews Autopilot?
Autopilot came out of Mews's acquisition of Atomize and continuously recalculates room rates based on demand, competition and projected occupancy. According to Mews, the platform runs more than 150 million pricing calculations a day, and 70% of former Atomize users now leave pricing running on full autopilot.
The claimed gain for a revenue manager is 20 to 30 hours a month, time previously spent manually adjusting prices room by room, channel by channel. Mews also cites one customer whose revenue grew 38% year on year after switching to Autopilot, a single case to treat as an illustration rather than a general promise.
What changes with self-service check-in through Mews Kiosk?
Kiosk digitises arrival: the guest scans their ID, collects their key and confirms payment in under 3 minutes, without going to the front desk. Mews reports that more than 60% of guests choose this route when it is offered, a signal that waiting is not the main reason guests value a physical front desk.
Another effect the vendor measures: upsell offers presented on the kiosk convert 2.6 times better than at the front desk. The most likely explanation is that the guest decides alone, without the social filter of a face-to-face exchange, a point worth checking against your own property before drawing a firm conclusion.
Does Mews offer an official AI agent or MCP server?
No, not to date. Unlike some other SaaS vendors that publish an official MCP server for connecting an external agent, Mews does not document any MCP server under its own name on its product or developer pages. There is also no official Mews chatbot for handling guest reviews or inbound requests.
What Mews does document is a fairly complete open API for building your own integration, at docs.mews.com.
How do you connect an external agent to Mews through its open API?
The Mews API breaks down into several pieces: the Connector API (access to bookings, rates, guest profiles), the Channel Manager API (multi-channel distribution), the Booking Engine API (the booking flow) and the Point-of-Sale API (on-site payments), complemented by webhooks that notify a third-party system in real time whenever an event occurs (a new booking, a check-in, a payment).
An AI agent does not connect to Mews natively: you need to build the bridge that translates its requests into calls against this API, exactly as for any third-party application installed on the platform. This is a conventional integration build, not an MCP connection in a few clicks.
What tasks could an external agent automate on Mews?
Once the bridge is in place, an agent connected to the webhooks and the Connector API can read upcoming bookings to prepare a daily occupancy summary, cross-reference data across several properties in the same group for consolidated reporting, or trigger a personalised pre-arrival message without touching the rate or the payment.
Write tasks (changing a rate, cancelling a booking, taking payment) stay the most sensitive: they are worth keeping under human validation until the agent's reliability is proven on the property in question.
How do you get started without taking on risk?
The sequence to favour starts with read-only access: reporting, alerts, preparing follow-ups, nothing that touches rates or the booking itself. That is the lowest-risk phase and the fastest to set up with the existing API.
Controlled write access comes next, on reversible and validated actions, never in bulk across several properties without oversight. Before investing in a custom Mews bridge, an assessment prices out the uses that will genuinely save time on your PMS and scopes the permissions the agent will receive.
Frequently asked questions
- Does Mews have an official AI agent or MCP server?
- No, not at the time of writing. Mews documents an open API (Connector API, Channel Manager API, Booking Engine API, Point-of-Sale API, webhooks), but the vendor has not published an official MCP server. An AI agent can still connect through that API, via a conventional integration build rather than a native MCP standard.
- What does Autopilot actually do inside Mews?
- Autopilot, built on the acquired Atomize technology, recalculates room rates continuously, running more than 150 million pricing calculations a day across the Mews platform. Among former Atomize customers, pricing now runs on full autopilot for 70% of users, with an estimated saving of 20 to 30 hours a month for a revenue manager.
- Does automated check-in replace the front desk?
- No, it absorbs the standard flow. Mews Kiosk enables check-in in under 3 minutes, and more than 60% of guests choose self-check-in when it is offered. The front desk stays necessary for special cases, specific requests and the guest relationship, automation just frees up time on the repetitive part.
- How can a hotel connect an AI agent to Mews without waiting for an official tool?
- By building on the open API Mews documents at docs.mews.com: a developer or integrator builds a bridge that queries the Connector API or receives webhooks in real time, then exposes that data to an AI agent with precise read permissions, and write access only on low-risk, approved actions.
Sources
Go further
AI in hospitality and restaurants