Connect an AI Assistant to TF Rooms
TF Rooms lets an AI assistant look up rooms, rates and prices at New Leaf Properties. It only reads: booking stays with the guest.
Server version: 1.2.1
Last Updated: November 6, 2026
What It Is
TF Rooms is a read-only MCP (Model Context Protocol) server for New Leaf Properties, twelve motels and extended-stay locations in Twin Falls, Idaho. In the protocol the server reports the title TF Rooms and the name newleaf-portfolio. An assistant that shows the title shows TF Rooms; others may show the name, or what you enter when you add it.
An assistant connected to it can answer questions such as which property has a room free on particular nights, what a room costs, and what one stay would total.
Connection Details
- Server URL:
https://tfrooms.com/mcp - Transport: Streamable HTTP. Each message is an HTTP POST and each answer is one JSON body. The address is for assistants: open it in a browser and it only says that it is POST only.
- Sign-in: None. There is no account, API key or OAuth step.
- Protocol version: The server reports
2025-06-18. It also accepts2025-03-26and2024-11-05when an assistant asks for one of them. - For developers: The server card and the full technical description list every limit.
How to Connect
Claude
- Open Settings, then Connectors.
- Choose Add custom connector.
- Paste
https://tfrooms.com/mcpas the URL. If it asks for a name, use TF Rooms. - Leave authentication empty. None is needed.
Other Assistants
Any assistant that accepts a remote MCP server URL can use the same address with no authentication. Menu names differ from one product to another and can change, so look for the place that adds a connector or a remote MCP server.
The Four Tools
Every tool only reads, and each one says so to the assistant in its annotations.
find_rooms
Asks all twelve properties, at the moment of the call, which have a room free tonight or on particular nights, and applies each property’s minimum stay. A property that does not answer is reported as unknown, not as full.
- Inputs:
checkinandcheckout(YYYY-MM-DD, both or neither; with neither it answers for tonight) androomtype(optional: single, double or studio). - Example: “Is there a double free anywhere in Twin Falls from December 1 to December 3?”
list_properties
Lists the twelve properties with name, address, phone, the room types each sells online, rates and booking page. It says nothing about whether a room is free.
- Inputs:
roomtype(optional: single, double or studio). - Example: “Which New Leaf properties have studios?”
get_rates
Gives what a room costs by the night and by the week, and the rules for a stay: the minimum stay, how a stay under a week is priced, the 30-day cap, the weekly-only studio and the 8% Idaho lodging tax.
- Inputs: None.
- Example: “What does a single cost by the week?”
quote_stay
Gives the total for one stay at one property. It asks that property’s own server and returns the answer word for word, including a refusal when a stay cannot be sold. It gives a price, not availability.
- Inputs:
property(a name, slug or domain, or a short name such as apollo; if a short name fits more than one property it asks which one),roomtype(single, double or studio),checkinandcheckout(YYYY-MM-DD). All four are required. - Example: “What would a single at the Apollo cost from December 1 to December 3?”
What It Never Does
TF Rooms does not book a room, take a payment, hold a room or ask for personal details.
Booking happens on each property’s own booking page (all twelve are on our locations page), where the guest signs the guest agreement and sends the form personally. Nothing is charged online; payment is made at the property.
Data Handling
- What a tool call receives: Only the tool’s inputs: dates, a room type and a property name. No tool has an input for a name, phone number, email address or payment details. The server does not receive your conversation with the assistant, only what the assistant sends for the tool.
- What the server keeps: Nothing. It uses the inputs to answer and has no database or file storage. A call that works writes no log line. Two failures do: when a property does not answer a quote, the server logs that property’s site name and the error, and when a tool fails, it logs the tool’s name and the error. Neither line records the dates or the room type.
- Where the inputs go:
find_roomssends the dates to all twelve property sites, in the web address of each request.quote_staysends the room type and dates to the one property named. Those sites store nothing from these requests and log only their own faults. - Hosting: The server and the property sites are hosted on Netlify, which processes every request as the hosting provider. That includes the network address a request comes from and, for
find_rooms, the web address with the dates. Netlify also holds the failure lines above. Our privacy policy describes the host’s server logs as an ordinary short retention cycle; we do not have an exact period to state. See Netlify’s privacy policy. - Accounts, cookies and tracking: The MCP endpoint has no accounts, sets no cookies and runs no analytics or tracking. That is about the endpoint. This web page is an ordinary page of tfrooms.com, so the analytics in our cookie policy apply here as they do on our other pages.
The privacy policy covers the connector in Section 16.
Limits and Accuracy
- Live answers: An answer is true at the moment of the call. There is no cached copy behind it, and a room free now can be taken a minute later.
- Quotes: A quote comes from the property’s own server, so it is the figure that property’s booking form uses. If the property does not answer, there is no figure, only its phone number and booking page.
- Booking links: A prefilled booking link reserves nothing. It opens the form with the stay filled in, and the guest still has to sign and send it.
Support
Email eric@tfrooms.com or call (208) 733-2010. The office is open Monday to Friday 8AM–6PM and Saturday 10AM–6PM, and closed Sunday.