How to manage your vacation rental from ChatGPT, Claude, or Gemini

How to manage your vacation rental from ChatGPT, Claude, or Gemini using Nowistay's MCP server

Anthropic released the Model Context Protocol (MCP) as an open standard in November 2024. OpenAI added native MCP support to ChatGPT in early 2025, Google followed with Gemini, and within a year MCP became the dominant way for AI assistants to securely access external tools and data. For vacation rental hosts, this is the moment a question like "what bookings do I have this week and which units need cleaning before Friday ?" stops requiring a dashboard, an inbox, and three browser tabs, and starts working as a single sentence to an AI assistant.

Nowistay was among the first vacation rental platforms to ship an MCP server. This guide explains what MCP is in plain terms, what you can actually do with it once it's connected, how to set it up in ChatGPT, Claude, or Gemini, and where the boundaries are around privacy, authentication, and which actions an AI assistant can take on your behalf.

What MCP is in one paragraph

MCP is an open protocol (created by Anthropic, adopted by OpenAI and Google) that lets AI assistants connect to external services in a structured, authenticated way. Before MCP, hooking ChatGPT or Claude up to your booking platform required custom plugins per platform, with different specs, auth flows, and capabilities. After MCP, any service that exposes an MCP server can be added to any MCP-capable AI assistant with a standard OAuth handshake. The AI assistant gets a defined set of tools (functions it can call) and resources (data it can read), under your account, with the permissions you grant.

For vacation rental platforms, MCP means a host can ask their AI assistant of choice natural-language questions about their portfolio and get real answers from live data, without leaving the chat. No dashboard, no inbox switching, no copy-paste of booking IDs into a search bar.

What you can actually do with Nowistay's MCP server

The MCP server exposes a curated set of tools and resources. Examples of natural-language queries hosts use most often:

Calendar and bookings

• "What bookings do I have this week ?"
• "Show me all bookings for Apartment 12B between June 10 and June 20."
• "Which of my properties has the lowest occupancy in July ?"
• "How many nights did I sell across the portfolio last month ?"

Operations

• "Which units need cleaning before Friday ?"
• "Are there any open team inquiries that haven't been answered in the last 6 hours ?"
• "List all guests checking in today, in order of arrival time."
• "What is the access code for the booking arriving tomorrow at the beach house ?"

Messaging

• "Show me the last conversation with the guest in Apartment 4A."
• "Are there any unread guest messages right now ?"
• "Summarize the messages I received overnight."

Reporting and analytics

• "What's my average review score across the portfolio for the last 30 days ?"
• "Which property has the highest revenue per available night this quarter ?"
• "How does this June compare to last June for total revenue ?"

The exact tools available depend on the version of the Nowistay MCP server and the permissions you grant during the OAuth handshake. Read-only queries are the default. Action-taking tools (like sending a message or modifying a booking) require an explicit additional permission scope and are clearly labeled.

How OAuth authentication works (in plain terms)

When you connect Nowistay to ChatGPT or Claude via MCP, the flow looks like this:

1. In your AI assistant, you add Nowistay as an MCP server and click connect.
2. A browser tab opens to app.nowistay.com and asks you to log in (or, if you're already logged in, to confirm the connection).
3. Nowistay shows you the permission scope: which tools the AI assistant is requesting, which resources it can access, and whether it can take any actions. You can decline or accept each scope.
4. If you accept, a secure token is issued to your AI assistant. The token is bound to your account and the specific scopes you granted.
5. From that point on, the AI assistant can call the granted tools on your behalf. Your password is never shared with the AI assistant. You can revoke the token at any time from your Nowistay account settings.

The auth flow is identical to how you connect any other service (Google Drive, Slack, GitHub) to an AI assistant. The security model is the same, with the OAuth provider (Nowistay) holding the credentials and issuing scoped tokens.

Setting up the Nowistay MCP server in each major AI assistant

ChatGPT (OpenAI)

In ChatGPT settings, open the Connectors panel and click "Add MCP server". Enter the Nowistay MCP server URL (provided in your Nowistay account settings under "Integrations > MCP"). Complete the OAuth handshake. The Nowistay tools become available in any ChatGPT conversation.

Claude (Anthropic)

In Claude desktop, open Settings > Developer > Edit Config. Add Nowistay to the mcpServers section with the server URL. Save and restart Claude. In Claude.ai (web), use the Connectors panel under Settings > Integrations and add Nowistay as an MCP source.

Gemini (Google)

In Gemini Advanced settings, open Integrations and click "Add MCP server". Provide the Nowistay MCP server URL and complete the OAuth handshake. The Nowistay tools become available in any Gemini conversation.

Other MCP clients

Any MCP-capable client (Cursor, Continue, custom MCP clients) can connect using the same server URL and OAuth handshake. The protocol is the same regardless of the client.

What the standard manual approach looks like

Without an MCP server, hosts who want this kind of access cobble together one of these:

• Direct dashboard access only. Log into the PMS web app or mobile app every time you want to check something. Works fine but breaks the workflow when you're in the middle of an AI-assisted task (planning a marketing campaign, drafting a financial report, comparing months).
• Manual data export. Export bookings or reports as CSV, paste into ChatGPT, ask questions about the static data. Works for one-off analysis but the data is stale within hours.
• Custom API integrations. Build a custom GPT or a Claude project that hits your PMS's API directly. Possible if your PMS exposes a public API and you're technical, but maintenance falls on you and the integration breaks every time the API changes.
• No integration. Just keep using the dashboard and accept the context-switching overhead.

How Nowistay built the MCP server

Nowistay's MCP server exposes a read-first set of tools covering bookings, properties, missions, messages, reviews, and reports. Action-taking tools (send a message, generate a smart lock code, create a cleaning mission) are gated behind separate permission scopes that the host approves during OAuth. The server runs under the same authentication and authorization model as the Nowistay web app: each user only sees the properties they own or manage, role-based permissions are enforced at the tool level, and audit logs record every MCP call. The server is hosted at a stable URL and updated with the rest of the platform. Hosts can disconnect or revoke access at any time from their account settings.

Whether you connect via Nowistay's MCP server, build a custom API integration against your existing PMS, or just stick to manual dashboard access, the criteria above (OAuth-based authentication, scoped tools, read-first defaults, audit logs) are the test for any AI-to-PMS connection that handles real booking data.

Privacy and access scope considerations

A few practical points hosts should know:

• Conversation history with the AI assistant. When you ask ChatGPT, Claude, or Gemini a question that involves your booking data, the data passes through the AI provider's servers as part of the conversation. Check your AI provider's data retention and training policies. OpenAI's enterprise tier, Anthropic's Claude for Teams, and Google's Workspace Gemini all offer policies that exclude your data from training.
• Permission scopes are granular. You can grant read-only access without granting any action-taking permission. Most hosts start with read-only and grant action scopes later if needed.
• Multi-account considerations. If you manage properties for multiple owners through a property management business, ensure your OAuth grant matches the scope of properties the AI assistant should see.
• Revoking access. Disconnect from Nowistay's account settings (Integrations > MCP > Connected Apps). The token is invalidated immediately and the AI assistant loses access.

Use cases that change the workflow

A few patterns hosts adopt within the first month of connecting an MCP server:

• Morning briefing. "Give me a summary of today: arrivals, departures, units that need cleaning, any open issues." Replaces three separate dashboard checks.
• Pricing review. "Compare last month's revenue per property to the same month last year, and flag the units that underperformed by more than 10%." Replaces a spreadsheet export and manual analysis.
• Guest history lookup. "Has the guest with email X.Y@example.com stayed before, and if so where ?" Replaces an inbox search.
• Investor reporting. "Draft a paragraph for my owner update on Apartment 12B's performance in Q2." Combines a data query with the AI assistant's writing capability.

What MCP does not replace

A few capabilities still belong on the Nowistay web app or mobile app:

• Initial property setup (uploading photos, configuring amenities, setting house rules).
• Direct booking page customization (branding, fonts, hero images).
• Billing and subscription management.
• Complex configuration changes (cleaning team assignment rules, payment splits, tax breakdowns).

MCP is best at queries, summaries, and structured data work. The dashboard is best at multi-step configuration and visual review. Hosts use both, depending on the task.

Getting started in 10 minutes

1. Log into your Nowistay account.
2. Open Settings > Integrations > MCP.
3. Copy the MCP server URL provided.
4. In your AI assistant of choice (ChatGPT, Claude, Gemini), open the integration or connector settings and add a new MCP server with that URL.
5. Complete the OAuth handshake, reviewing each permission scope.
6. Ask your AI assistant: "What bookings do I have this week ?" The answer should come back in seconds.

From there, the workflow is open. The hosts who get the most value are typically those who use AI assistants regularly for other tasks (writing, analysis, planning) and who appreciate the friction reduction of having their portfolio data available in the same conversation.