# Cloudbeds MCP server for front desk teams

> A Cloudbeds MCP server for front desk teams: let staff look up guests and reservations in Claude or ChatGPT while rates and payments stay off their role.

**TL;DR** A Cloudbeds MCP server for front desk teams lets staff ask Claude, ChatGPT, Cursor or Microsoft Copilot about arrivals, guests and housekeeping. Elaichi's Cloudbeds connector carries 490 tools, and a restriction on the front desk role keeps rate, payment and guest-erasure tools out of reach. Each agent signs in with their own Cloudbeds login, and the audit log shows who ran which tool.

## Can a Cloudbeds MCP server for front desk teams keep rates and payments off the desk?

Yes. Connect Cloudbeds in Elaichi with OAuth, add https://api.elaichi.ai/mcp in Claude, ChatGPT, Cursor or Microsoft Copilot, and hold the front desk role to ten named desk tools with a restriction on the role. Rate, payment and guest-erasure tools then stay out of reach.

A front desk shift is a run of small questions. Who arrives before noon, which rooms are clean, whether the guest in 212 has a note about a late checkout. MCP (Model Context Protocol) is the standard way an AI assistant calls tools in other apps. Cloudbeds is a property management system, the software a hotel runs its reservations and front office on (see [what a property management system does](https://en.wikipedia.org/wiki/Property_management_system), checked November 2026). The [Cloudbeds connector](/connectors/cloudbeds/) lets the assistant read and change those bookings.

## What do Cloudbeds keys and roles control?

Keys carry scopes chosen at creation. Roles decide what each user may do inside the app. The two are separate controls.

For keys, the person creating one picks scopes from a list that should match the endpoints used. Changing scopes means deleting the key and making a new one. A key is shown once, and it stays alive while used at least every 30 days (checked November 2026, [Cloudbeds quickstart for API authentication](https://developers.cloudbeds.com/docs/quickstart-guide-api-authentication-for-property-level-users)). That page describes scopes and does not describe role privileges.

For roles, Cloudbeds groups privileges into rates, financial reports, guest data and payments, among others. Sensitive guest fields are hidden by default, and merging or anonymizing a guest are privileges of their own (checked November 2026, [Cloudbeds role privileges](https://myfrontdesk.cloudbeds.com/hc/en-us/articles/19300997644443)).

## Whose Cloudbeds login should run the desk's calls?

Each agent's own. In Elaichi, every call through a connection uses the account that connection was made with, so one connection per agent ties each call to one person.

Connect with OAuth. The agent signs in to Cloudbeds themselves, and OAuth hands Elaichi a token with limited scope instead of their password (see [RFC 6749](https://www.rfc-editor.org/rfc/rfc6749), checked November 2026). A shared connection would run every agent's call as whoever connected it, and Cloudbeds would see one user. An API key is made by a login that can reach the API Credentials page, and it is one secret for whoever holds it. Cloudbeds' docs do not say a key inherits its creator's role, so the restriction below does not lean on that. After an admin edits a role in Elaichi, the change takes effect in about two minutes.

## Which Cloudbeds tools should the desk reach?

Ten: the lookups, the stay actions and the notes. Rates, payments and erasure belong to other roles.

The connector carries 490 tools, 49 of them deletes. For arrivals and guests, the desk needs list_all_cloudbeds_reservations, get_single_cloudbeds_reservation_by_id, cloudbeds_guests_search and list_all_cloudbeds_guest_notes. For the stay: cloudbeds_reservations_check_in, cloudbeds_reservations_check_out, cloudbeds_rooms_assign and create_a_cloudbeds_reservation_note. For the floor: list_all_cloudbeds_housekeeping_statuses and list_all_cloudbeds_available_room_types.

Keep these off the desk role:

- cloudbeds_rates_bulk_update and create_a_cloudbeds_rate_plan change what every future guest pays.
- create_a_cloudbeds_payment, create_a_cloudbeds_charge, cloudbeds_payments_void and cloudbeds_folio_items_void move money.
- cloudbeds_guest_profiles_anonymize replaces personal fields with placeholders.
- create_a_cloudbeds_door_lock_key issues a physical key.

## Should the desk rule be an allow rule or block rules?

An allow rule. More than half of Cloudbeds' 490 tools are not reads, so block rules would have to name hundreds of tools, and the desk needs only ten.

In Elaichi, restriction targets are role or user only; the organization default is the absence of a rule, which means allow everything. Holding the desk to its ten tools takes an allow rule naming them. The catch is that once a role carries any allow rule, every connector and tool none of its allow rules names is blocked for that role. If the desk also uses Gmail or a chat app through the same endpoint, add allow rules for those. Block rules suit the opposite case, because a layer with only block rules has no allowlist, so everything it does not block passes it: they work when the non-read tools are few enough to list. [Per-tool and per-app restrictions](/blog/per-tool-vs-per-app-restrictions/) covers the trade-off.

## What happens when the desk asks for a tool it lacks, and what is logged?

The agent gets a clear answer, and the manager gets a record of what did run. A restricted tool is left out of the tools Elaichi offers a client and cannot be called. Search still names it, flagged restricted, with no schema.

Elaichi writes one entry for each connected-tool call that reaches execution (a call refused earlier writes none). After a dispute, a manager can see which agent's grant ran cloudbeds_reservations_check_out, on which connection, from which AI client. Elaichi keeps 90 days of that history. For a longer record, export to your own Datadog comes with the Black plan.

## Is one revenue manager's own Cloudbeds key enough?

Yes, when the revenue manager is the only person who will use Cloudbeds in an AI client. A key with the scopes they pick, connected by them alone, needs no desk role and no second person to attribute calls to.

The setup in this post earns its keep when the desk, the revenue manager and housekeeping leads share one hotel's data and need different slices. With one property and one person doing every job, it is more setup than the work needs. The [Cloudbeds connector page](/connectors/cloudbeds/) lists every tool by name, and the [whole connector catalog](/connectors/) shows the other apps a hotel team may add.

## FAQ

### Can a front desk agent check guests in through an AI assistant without touching rates?

Yes. In Elaichi, a restriction on the front desk role can allow cloudbeds_reservations_check_in, cloudbeds_rooms_assign and the reservation lookups, and name nothing else from Cloudbeds. Rate tools such as cloudbeds_rates_bulk_update then stay restricted for that role, so the assistant cannot call them even if the agent asks.

### Do Cloudbeds role privileges apply to calls made through Elaichi?

Only as far as Cloudbeds applies them to the credential the call uses. In Elaichi, every call through a connection runs on the account that connection was made with. Cloudbeds defines rate, payment and guest-data privileges per role (checked November 2026, https://myfrontdesk.cloudbeds.com/hc/en-us/articles/19300997644443), so connect each person with their own login and cap the AI path with a restriction as well.

### How do I set up Cloudbeds with Claude or ChatGPT?

Each agent connects Cloudbeds in Elaichi with their own login through OAuth, then adds https://api.elaichi.ai/mcp as a connector in Claude, ChatGPT, Cursor or Microsoft Copilot and approves their own grant. The model finds Cloudbeds tools with search_tools and runs them with execute_tool, so the client never lists 490 tool names.

### What does the audit log show for a Cloudbeds call?

In Elaichi, the audit log holds one entry for each connected-tool call that reaches execution (a call refused earlier writes none). Each entry names the person, the tool, the connector, the connection and the AI client. The log records what ran, not everything attempted. Elaichi keeps 90 days of audit history.

## Read next

- [Restrict one AI tool or the whole app? Six cases](/blog/per-tool-vs-per-app-restrictions/) — Restrict one AI tool when a role needs part of an app, and block the whole app when it needs none of it. Six cases, and what new tools do to each rule.
- [Read-only ChatGPT access to your company apps](/blog/make-chatgpt-read-only-in-company-tools/) — Read-only ChatGPT access comes from a restriction rule on a role, not from a consent checkbox or a client setting that sounds like one.
- [Share an MCP connection without the API key](/blog/share-an-mcp-connection-without-the-api-key/) — How to share an MCP connection without the API key in Elaichi: grant a teammate use of a connected account, and take that access back.
