# MyBooking.ai

> MyBooking.ai — AI operating system for hospitality and service businesses, with a marketplace for stays, real estate, flights and jobs.

MyBooking.ai is a business platform: one account to run hotels and apartments, guest communication, documents and accounting, tasks and teams, jobs, real estate, flights, medical appointments, transport, restaurants, legal cases and AI models.

MyBooking.ai (www.mybooking.ai) is not affiliated with Booking.com or with other services that use the name MyBooking, such as mybooking.es.

MyBooking.ai works with AI assistants (Claude, ChatGPT, Grok and any MCP client) through one MCP server. The assistant prepares actions; by default the user approves every change and every payment on MyBooking.ai. Prices are the same as on the website.

## Connect an AI assistant

- MCP server (Streamable HTTP, OAuth sign-in with dynamic client registration): https://www.mybooking.ai/mcp
- [Setup guide](https://www.mybooking.ai/en/ai/): how to connect Claude, ChatGPT or Grok
- Claude: Settings → Connectors → Add custom connector → paste the MCP server address → sign in to MyBooking
- ChatGPT, Grok and other clients: add the same address as a custom MCP connector

## Rules for AI assistants

- Always tell the user that the action goes through MyBooking.ai.
- Read-only tools run immediately.
- Tools that change data return a draft with status 'pending_approval' and an approval_url: give the user that link; nothing changes until they approve it (unless their MyBooking autonomy settings allow auto-execution).
- If the status is 'needs_input', ask the user for the 'missing_fields' and call the tool again with the complete payload.
- Before a paid action, tell the user the price shown in the tool result.
- Tools see only the data the signed-in user may access. Never ask the user for passwords or card numbers.
- If no listed tool can do what the user asks, say so and point them to https://www.mybooking.ai; new tools appear after the user removes and re-adds the MyBooking connector.
- If the user has no MyBooking account, call account_signup with their name and e-mail — never ask for a password: they receive an activation e-mail and a set-password link. Then ask them to sign in through this connector so you can act on their behalf.

## Actions available to AI assistants now

- `account_signup`: Create a MyBooking account for a person who has none. Never ask for a password: the person receives an activation e-mail and a set-password link. After activation they sign in through this connector. Requires explicit acceptance of the terms (accept_terms=true) (creates a draft that takes effect after the user approves it)
- `jobs_apply_draft`: Apply to a job offer on behalf of the logged-in user (employee or B2B contractor lane is detected automatically). Uses the user's profile; missing profile fields can be passed in. Daily limit and duplicate checks are the same as on the website (creates a draft that takes effect after the user approves it)
- `jobs_close`: Close one of the user's job offers: it disappears from the public job board and stops taking applications. The offer and its applications stay in the user's account (creates a draft that takes effect after the user approves it)
- `jobs_my_offers`: List job offers posted by the user with status and number of applications (read-only, runs immediately)
- `jobs_post_draft`: Create a job offer for the user's business (hotel, cleaning, reception, IT…). Published on the MyBooking job board after the user's approval. Owner/manager accounts only (creates a draft that takes effect after the user approves it)
- `jobs_search`: Search public, published, non-expired job offers on MyBooking (hotels, cleaning, reception, IT, etc.). No login needed. Returns offer slugs to use with jobs.apply_draft (read-only, runs immediately)
- `knowledge_get`: Return the full text of one item from knowledge.search: a CourtExpert matter (court journal, filings, list of its documents), a document (converted text of all its files), one document file, an email (headers, body and the text of its attachments), a Messages conversation (all messages, chronological) or a phone call (full transcript + AI receptionist summary). Paged with offset/max_chars. Only items the signed-in user can see on the MyBooking screens (read-only, runs immediately)
- `knowledge_search`: Search the signed-in user's own knowledge in MyBooking with one call: CourtExpert legal cases (matters: number, title, court journal — the newest are listed even without a match, so 'the latest criminal case' can be resolved from the list), documents (uploaded files, converted text; hybrid semantic + full-text, grouped per document with the best fragments and the file list) emails (bodies + attachments), Messages conversations and phone calls (recording transcripts + AI receptionist sessions; `recent_calls` lists the newest calls when the question is about calls); `recent` lists the newest documents whose title matches, for questions like 'the latest … case'. Same visibility as the MyBooking screens. Pass matter_number to restrict documents to one case. Then use knowledge.get to read a matter, a whole document, a file or an email in full (read-only, runs immediately)
- `machines_list`: List the user's computers enrolled in MyBooking GPU Fabric: type, online or offline, what the agent is doing (ready, busy, paused and why), battery, free disk and memory, and the work mode (when the machine may take tasks). when_idle = new tasks only when nobody has used the computer for idle_minutes (default 10), a started task finishes in the background; always = whenever the agent runs; night = only between night_from and night_to (the machine's local hour); manual = paused (read-only, runs immediately)
- `machines_set_mode`: Change when one of the user's GPU Fabric machines may start new tasks: when the computer is not used, always, only at night, or paused; and whether a laptop may start tasks on battery. The agent on the machine picks it up within a minute. when_idle = new tasks only when nobody has used the computer for idle_minutes (default 10), a started task finishes in the background; always = whenever the agent runs; night = only between night_from and night_to (the machine's local hour); manual = paused (creates a draft that takes effect after the user approves it)
- `mail_detect_settings`: Detect IMAP/SMTP servers and ports for an e-mail address or domain (known providers, MyBooking mail server, MX records, connection probe, AI fallback). Returns the settings, the source and a confidence level plus a note (e.g. 'use an App password'). No login needed to the mailbox (read-only, runs immediately)
- `mail_import_now`: Start an import of new messages from a connected mailbox right away (one job per folder), instead of waiting for the scheduled run. Needs the password set. Draft, needs approval (creates a draft that takes effect after the user approves it)
- `mail_mailbox_add`: Connect an e-mail mailbox (IMAP/SMTP) to the user's MyBooking account so its messages are imported and searchable. Servers are detected from the domain. The PASSWORD IS NEVER PASSED HERE: after approval the result gives a connect_url on MyBooking.ai where the user types the password; the connection is tested there before anything is saved and the import starts right after. Optional company_id sets the address as that company's e-mail — ask the user first whether the mailbox belongs to an existing company. Draft, needs approval (creates a draft that takes effect after the user approves it)
- `mail_mailbox_remove`: Disconnect an extra mailbox from the user's account: its settings and password are deleted and the import stops. Already imported messages are no longer shown. The main mailbox cannot be removed this way (change its settings instead). Draft, needs approval (creates a draft that takes effect after the user approves it)
- `mail_mailbox_test`: Log in to a connected mailbox over IMAP with the stored password and list its folders. Tells whether the password is set and whether the login works. Read-only (read-only, runs immediately)
- `mail_mailbox_update`: Change the servers, folders, automatic import or interval of a connected mailbox. To change the PASSWORD, send the user to the mailbox's connect_url (mail.mailboxes); it is never passed here. Draft, needs approval (creates a draft that takes effect after the user approves it)
- `mail_mailboxes`: List the e-mail mailboxes connected to the user's MyBooking account (main + extra): servers, import status, number of imported messages, last sync. Passwords are never returned. Also lists the user's companies with their e-mail address (read-only, runs immediately)
- `mail_share`: Share one of the user's mailboxes with another user of their team: an owner can share with their own employees, an employee with their owner (mail.shares -> candidates lists who). Reading is always granted; can_send adds sending (default true). Calling it again for the same user updates can_send. Draft, needs approval (creates a draft that takes effect after the user approves it)
- `mail_shares`: List the shares the user granted on their mailboxes, the mailboxes shared with them, and the users they may share with (owner: their own employees; employee: their owner) (read-only, runs immediately)
- `mail_unshare`: Remove a share the user granted on one of their mailboxes (share_id from mail.shares). Draft, needs approval (creates a draft that takes effect after the user approves it)
- `market_rates`: Competitor prices from Market Intel for an arrival day: per competitor set the latest capture for that day, statistics (min, p25, median, p75, max) and competitors sorted from cheapest with room type, city and district (filter by district yourself: sets can be city-wide). Own properties excluded (read-only, runs immediately)
- `phonebook_caller_name`: Name of a phone number in the phone book (contact or team member) (read-only, runs immediately)
- `phonebook_create`: Add a contact (Team by default) with one or more phone numbers (creates a draft that takes effect after the user approves it)
- `phonebook_import`: Import contacts (Private by default). With dry_run, returns the preview with duplicates marked (creates a draft that takes effect after the user approves it)
- `phonebook_search`: Search the phone book (own contacts, Team contacts, team members) by name, company, e-mail or number (read-only, runs immediately)
- `phonebook_update`: Change name, numbers, e-mail, company, notes or Team/Private of a contact (creates a draft that takes effect after the user approves it)
- `property_create_draft`: Create a new property for the user with its rooms. Creates the MyBooking sales channel automatically. The platform fee for adding a property is charged from the user's token wallet (same as the web form). Publication to other channels is a separate action (creates a draft that takes effect after the user approves it; paid from the user's token wallet)
- `property_list`: List the properties (hotels, apartments, buildings) owned by the user, with room counts (read-only, runs immediately)
- `rates_get`: Prices and availability MyBooking sends to each sales channel (Booking.com via Channex ...), per offer and date: price, currency, rooms to sell, closed flag and sync status, plus the connection's price floor rules. Read this before proposing price changes (read-only, runs immediately)
- `rates_recommend`: Ask the pricing assistant built into MyBooking (the /pricing-ai/ model, with your offers, prices and sales history) to propose prices. Paid from the MyBooking token wallet — the approval card shows the estimate. Returns a task_id: read the result with rates.recommendation (30-120 s), then apply it with rates.set(from_recommendation=task_id), all or only_offers (creates a draft that takes effect after the user approves it)
- `rates_recommendation`: Result of rates.recommend: phase (planning / dryrun / done / error), summary, per-offer proposals (offer id, name, current, proposed, rationale) or the plan actions in plain words, errors and the real cost. Apply it with rates.set(from_recommendation=task_id) (read-only, runs immediately)
- `rates_set`: Change prices of the user's offers, through the same engine as the MyBooking pricing assistant (/pricing-ai/): explicit prices per offer and night range, or the plan of a rates.recommend recommendation (from_recommendation). The approval card shows the effect night by night, channel floors and the safety check (> 10 offers or below 60% of the ADR needs confirm_count = number of offers). After approval the prices are saved as one batch (undo with rates.undo or on /pricing-ai/) and pushed to the channels (creates a draft that takes effect after the user approves it)
- `rates_undo`: Undo one applied price batch (rates.set or the /pricing-ai/ screen): created rules are switched off, manual prices removed, changed rows get their old values back, then the channels are updated. Same as 'Undo batch' on /pricing-ai/ (creates a draft that takes effect after the user approves it)
- `realestate_create_draft`: Create a real-estate listing for sale or rent in the user's brokerage. Saved as a draft; use realestate.publish to make it public. Requires an active agent profile (the tool explains how to set one up if missing) (creates a draft that takes effect after the user approves it)
- `realestate_my_listings`: List the real-estate listings of the user's brokerage (requires an active agent profile) (read-only, runs immediately)
- `realestate_publish`: Publish a draft listing on the MyBooking marketplace (status active). Accounts created via AI must have approved identity verification (KYC) first (creates a draft that takes effect after the user approves it)
- `realestate_search`: Search public, active real-estate listings on MyBooking (sale, long/short rent, lease, auction). No login needed. Returns listing slugs/references and public links (read-only, runs immediately)
- `realestate_unpublish`: Withdraw one of the brokerage's listings from the MyBooking marketplace (status withdrawn). The listing stays in the broker dashboard and can be published again with realestate.publish (creates a draft that takes effect after the user approves it)
- `reservation_get`: One reservation by channel number (every room of a multi-room booking: -0, -1 ...) or internal id: everything reservation.list returns plus cancellation policy, commission, extra beds, fees, invoice request and check-in progress. Guest contact data, card data and access codes are never returned (read-only, runs immediately)
- `reservation_list`: List reservations of the user's properties like the Daily Report screen: guests staying in a date range, arrivals, departures or bookings made. Each item: canonical status (confirmed = a room is assigned, cancelled, unknown), raw channel status, payment hint, dates, nights, guests, amounts, offer and listing, building, assigned physical rooms, flags (no_room_assigned, zero_amount). Guest data is masked to initials. Plus a summary with counts (read-only, runs immediately)
- `rooms_occupancy`: For one night and each of the user's buildings: physical rooms in use (overnight booking mode, active), rooms occupied by a non-cancelled reservation (RoomBooking), free room numbers, and active reservations of that building that have no room assigned yet. Use it before accepting, moving or pricing reservations: the real capacity is physical rooms, offers share them (read-only, runs immediately)

## Platform modules

- Accommodation: properties and rooms, bookings, PMS, channel manager for Booking.com and other OTAs, dynamic pricing and the MyBooking sales channel
- Communication: guest messaging, VoIP telephony with an AI receptionist and a shared phone book
- Documents: document storage with AI naming and search, and invoices captured from e-mail
- Accounting: bookkeeping, sales and cost invoices, and Polish KSeF e-invoicing
- Task management: tasks for staff and contractors, such as cleaning, reception and maintenance, billed per company
- Talent Hub: job offers and job search
- Real estate: listings for sale and for rent
- Flights: flight search and booking, charters and empty-leg flights
- Medical: doctors and clinics with online appointment booking
- Transport: rides and airport transfers
- Restaurants: table reservations, menus, kitchen display and waiter app
- CourtExpert: legal case management with matters, deadlines, hearings and AI-drafted filings
- AI models: model training and federated learning

Modules without a listed action are used on the website.

## Links

- [Website](https://www.mybooking.ai/en/)
- [AI assistants guide](https://www.mybooking.ai/en/ai/)
- [Terms](https://www.mybooking.ai/en/terms/)
- [Privacy policy](https://www.mybooking.ai/privacy/)
- Support: support@mybooking.ai
- Operator: Abba Villa Host Central Warsaw sp. z o.o.