Anny AI

Research published traveller reviews for the hotel and its Comp Set. Each platform is reported separately with the rating on its own scale, the review count, and when it was read; ratings are never combined across platforms. Themes come with short dated quotations and sourceIds.

comp_set_reviews_get

Correct the facts of one tracked event: its name, dates, hotel-local start time, venue, location, or demand. Pass only what the user asked to change; null clears a fact. The corrected facts stay through every later combination. A combined listing cannot be edited: separate it first with tracked_event_separate.

tracked_event_update

Have Anny AI draft an operating document for one hotel: an SOP, policy, handbook, onboarding guide, training checklist, job description, the hotel's operating profile or another document the team works from. The operating profile, type operating_profile, is the hotel's "How <hotel> works" document: its systems, departments and roles, terms and voice, learned from every document the hotel can open and its hotel details; every later draft reads it first, and publishing a new one updates the hotel's profile. Anny AI writes it in the background from the request, the hotel's saved details and the team's own documents, and lists what it could not confirm as questions. It saves a draft that the team opens in the editor; Chat and MCP never answer from a draft until someone publishes it. It works with no documents at all. Put the person's request in instructions in their own words, with every fact they gave. Pass sourceDocumentIds only for documents they named as the basis.

document_draft

Research the places that bring people to this hotel’s market: employers, venues, hospitals, campuses, sports, retail, and transport hubs. Each driver carries its location, a distance when one is published, and sourceIds. It reports no demand, occupancy, or revenue numbers. For the hotel’s stored list that event searches use, call demand_drivers_list instead.

market_demand_drivers_get

Get the current Comp Set for a managed hotel. yourHotel is the user's own hotel and competitors are the hotels it is compared against, each with public name, address, phone, and website. teamEdited names the fields the user's team corrected.

comp_set_get

Get editable Comp Set state, with paginated nearby candidates (5 by default, maximum 20). Follow pagination.nearby.nextOffset with the same limit.

comp_set_editor_get

Rebuild one Anny AI dashboard screen with the same calculations its widgets run. today: what the Today screen shows for each selected hotel: briefs, each role's saved headline (hotel_brief_get reads one role's observations, actions and citations), executiveSummary when several hotels are selected, and outlook, the next 7 nights with your rate, the Comp Set median, when each was collected, and events. overview: change counts of hotel-and-date pairs (brandHigher counts dates where the hotel's own website charged more than an OTA, shown as Brand priced higher; brandLowest counts dates where its own website was the lowest price, shown as Brand priced lowest), the market snapshot, OTA price gaps, and for the week shown, each hotel's position (its average rate against its Comp Set median) and the largest competitor move. year: the twelve-month Pulse year. table: every hotel and competitor price with its change since the previous shop. chart: the line values. location: mapped hotels, tracked events and, when venues is true, the saved venues on the map (up to 100 per hotel, so page a few hotels at a time). Every tab but today also returns hotelStrip, the card of every hotel the user can see, selected or not. Omit managedHotelIds for every hotel. endDate is the day after the last arrival date, as in the dashboard link, and today always covers the next 7 stay dates. Quote these figures for questions about the dashboard instead of calculating your own. For a question about one hotel and stay date, such as its Comp Set movement or median, pass that hotel in managedHotelIds and selected: the selected result then lists each competitor's price, previous price, change, and when and how it was collected, which is the evidence behind that figure. When those prices were collected more than a day ago, end by offering a fresh shop of that date. When changedSinceView names a hotel, its rates were published again after the screen loaded: say the screen is out of date and give the current figure. Page per-hotel sections, hotelStrip included, with offset and limit; portfolio figures cover every selected hotel. A hotel id the user cannot see is left out and listed in view.unknownHotelIds. The Evidence tab is not rebuilt here: read hotel_evidence_get for its hotel and dates. On today, yesterday is each hotel's closed result for last night against the same weekday last year: answer how a past night actually ended from it. A brief's finishedCallouts are recent stay dates its briefs called out that closed near full, picked up strongly or got an action item, each with occupancy on the books at the first callout, final occupancy, and the action items made from it. Outlook nights add rooms on the books, ADR, pickup and last year's result, and pickup lists stay dates by their largest pickupDays change, 20 at most. A hotel without these has no imported results the user can read.

dashboard_view_get

Get one visible managed hotel.

hotel_get

Read the facts this workspace keeps for its hotel and the competitors in its Comp Set: check-in and check-out time, neighborhood, address, phone, website, description, and custom details such as parking or pet policy. Each value says whether it comes from the user's team or from Google Hotels. Times are the hotel's local wall-clock time, never UTC. Answer questions about these facts from this tool. For a rule, fee or policy at the user's own hotel, such as a pet fee, call documents_search first when it is offered: the team's documents outrank Google Hotels and the web. Team values and notes are private: never put them in a web search, apart from a corrected address, phone or website.

hotel_details_get

Get the current rate matrix. Each data collection returns up to 5 rows (the default). Follow pagination nextOffset with the same limit; exhausted collections return empty arrays. Dates must be paired, ordered and within 31 days (exclusive end). Use detailOffset/detailLimit to page nested rates and hotel channels via ratesPage/channelsPage; detailLimit defaults to 5, maximum 5. Use rate_summary_get for an overview.

rate_matrix_get

Get current rates, estimated demand, freshness confidence, and movement for each Comp Set hotel and arrival date. Omit dates to use the configured horizon. comparison gives the stay length compared, the first and last arrival date and how many there are, the user's hotel rate, and the Comp Set median for each date; quote it instead of calculating your own averages or counting dates. In Anny Chat the result is drawn as a chart and a table that labels and explains each price's rate type: call this again with chartSeries set to each-competitor when the user wants a line for every competitor instead of the Comp Set median. To show what one shop collected, pass its shopRunId with its first and last arrival date.

rate_summary_get

Get rate and estimated-demand history for one arrival date and length of stay, newest shops first. Each collection returns up to 5 rows (the default). Follow pagination nextOffset with the same arrivalDate and limit. Page nested rates/channels with detailOffset/detailLimit (default 5, maximum 5) using ratesPage/channelsPage. Exhausted collections return empty arrays.

rate_trend_get

Get latest shop results for a visible hotel.

shop_results_get

Get shopping status and schedule for a visible hotel. schedule.state is the only source for whether the hotel shops on a recurring schedule; rules are its configured shop days, each with its own next scheduled shop, and do not run while the state is not_scheduled. lastRun.startedBy says whether the latest shop came from the schedule or was a one-off run.

shop_status_get

Get one published Anny AI support article by slug. Set absoluteLinks to true when the Markdown will be read outside Anny AI.

knowledge_get

List a hotel's action items with their owners, due dates, and status, the teammates who can be given one, the hotel's own today, and a library of common action items for each hotel role. An item added from a daily brief point has briefPoint: the point's text, its brief generationId and role, the findingIds, eventIds and rateSourceIds it cited, the itemIds of the evidence items it cited, and its stay dates from startDate to endDate. To open that evidence, read hotel_evidence_get for the hotel and those dates with that generationId. startDate is null when the point named no stay dates. Read this before suggesting a plan or changing an action.

hotel_actions_list

List the places stored as demand drivers for a hotel: venues, attractions, campuses, airports, hospitals, employers, and recurring events within 40 miles, each with its tier, type, distance in miles, and a one-line reason, highest tier and nearest first. It is the list searches for events near the hotel use, with the hotel’s own corrections applied: places it hid, added, or re-tiered. counts gives how many places each view holds. Narrow with view (a tier, or hidden) and query (part of a name), and follow pagination.nextOffset with the same limit. Read it before demand_driver_update so the placeId is real.

demand_drivers_list

List hotels visible to the connected user in an organization.

hotels_list

List the common documents one hotel has and the ones it is missing: the operating documents most hotels keep, such as SOPs, policies and handbooks, by department (front office, housekeeping, maintenance, food and beverage, sales, revenue, safety and security, people). A document counts as present when one the hotel can open, its own or shared with the workspace, has the common name or a usual variant in its title. Each present document links the matching document. Read it when someone asks which documents they are missing or should have.

documents_missing

List organizations the connected user currently belongs to.

organizations_list

List recent rate reports visible in an organization.

reports_list

List the shop credits and pools that can pay for an on-demand shop run.

shop_pools_list

List recent shop runs for a visible hotel.

shop_runs_list

List the team's documents the user can open, newest first, 50 a page: title, hotel, whether the whole workspace can open it, whether Anny AI can answer from it yet, who added it, when, and its path. A document written in Anny AI also has wordPath and pdfPath, which download it as a Word file or a PDF. Follow nextPage for more.

documents_list

List the events a hotel is tracking, in date order, with hotel-local dates and start times (never UTC). Each event lists combinedListings: other listings of the same event that Anny AI combined into it. factsSetByTeam names the facts a person added or edited, which combining never changes. Narrow with from and to (hotel-local dates; an event still running on from is included), and follow nextCursor for more. Read it before changing a tracked event so the savedEventId and listingId are real. A saved event search is not a tracked event: for those, use event_searches_list.

tracked_events_list

Read a saved daily brief for one hotel and role, with citation IDs and source freshness. For supporting evidence, call again with evidence containing the returned generationId, citation kind, and one citation ID. Evidence reads return only the selected records for a finding, event, or rate source. An item is the exact record or event that an observation cites. If the generation changed, read the brief again. For every finding and event on chosen stay dates, including those the brief did not cite, use hotel_evidence_get. Uses the same authorized results as Today. Does not generate or refresh a brief. A missing brief is unavailable, not evidence of no changes.

hotel_brief_get

Read one public page as evidence, such as an official calendar, event detail, or hotel page. Returns the page text, links, and structured data with a sourceId to cite in square brackets. A blocked or unreadable page verifies nothing.

read_page

Read the text Anny AI extracted from one team document, 20,000 characters a call from offset. Follow nextOffset to read on. text is null while Anny AI is still reading the file or could not read it.

document_get

Read every evidence item for one hotel and its arrival dates from startDate to endDate (inclusive, at most 31 days): the typed rate findings and tracked events a brief is written from, not only those a saved brief cited. Read it to write a fresh analysis or brief for a hotel and dates, or to find what a saved brief left out, then answer from these items. Each item has its id, type, stay dates (start, end), leadTimeDays, values (each type's units are in types), missing (values that are null), comparison (its basis, and whether a prior value existed), observed (when each side was collected), and peers (each competitor's previous and current rate) when present. sources says for each source whether data was collected (available or missing), when, and how many items matched; a stay date without items has no finding. types also gives each type's limitations, the claims its values cannot support: respect them without restating them. Without generationId the evidence is current, rebuilt from the latest data, and generationId is null. With the generationId from hotel_brief_get it is the evidence that saved brief saw, and selectedFor names the roles whose writer received each item: an item with no role was left out of that brief. The same id in both reads is the same evidence; an id only in the current read is newer than the brief. Items come in date order, limit per page (10 unless set, at most 20): page.total counts every match, and a non-null page.nextOffset is the offset of the next page, so read on until it is null to see everything. Pass page.snapshot from the first page with each later offset: when page.changed is true the evidence changed between pages, so read it again from offset 0. Stay dates, businessDate and leadTimeDays are the hotel's local calendar: never convert them or label them UTC. Name each figure's stay date, hotel and collection time so it can be traced. Does not start a shop or a brief. For a hotel with imported PMS results, pms_ items add on-the-books rooms, pickup, availability, group, operations and closed-result findings, and sources.pms says when they were imported.

hotel_evidence_get

Read imported PMS data for one hotel and 1–31 dates (inclusive). Choose history for closed actuals, forecast for latest on-books snapshots with exact 1-, 3-, and 7-day room pickup, or market-mix for latest category forecasts. Includes stored room inventory, availability, occupied and sold rooms, ADR, revenue, RevPAR, and sourceFields with arrivals, departures and group allocations when reported. Occupancy is percent, money uses each row currency. Preserve nulls and missing dates, report settings and source freshness. Forecasts, including past Market Mix dates, are not closed actuals or Anny AI Estimated Demand. comparison holds closed actuals from exactly 364 days earlier (weekday-aligned SPIT), never historical on-books pickup. Empty arrays mean no imported rows in that scope. Travel Agents is an aggregate category, not named-agency production.

pms_data_get

Search published Anny AI support articles.

knowledge_search

Search the text of the team's own documents the user can open, such as policies, procedures, standards, handbooks and onboarding guides. Call it before hotel details research or web_search for any rule, fee or policy at the user's own hotel, even when the question names no document. Returns the best passage of each matching document with its title, heading, path and documentId; read more of one with document_get, and do so before saying a document does not answer part of the question. When a passage states a check-in or check-out time of the user's own hotel, the result's nextSteps say whether the team's details lack or give it: follow them before answering. Outside Anny Chat there are no nextSteps and no hotel_details_propose. When a passage states the phone, address or website of the user's own hotel, call hotel_details_get, and when it lacks or differs, finish with hotel_details_propose if offered. With managedHotelId, only that hotel's documents and the ones shared with the whole workspace.

documents_search

Search the public web for current facts about public places, hotels, venues, and events. Returns results with sourceIds; cite a result by writing its sourceId in square brackets, such as [1a2b3c4d]. Only public identity such as names, addresses, and websites may appear in a query: never rates, revenue, guest details, staff notes, or internal IDs. Not needed for greetings, ordinary conversation, or Anny AI product help.

web_search

Send an App Message from the user to teammates in Anny AI: people by memberId, Hotel Access Groups by groupId, or admins true for every Owner and Admin, with team documents attached by documentIds. Teammates who cannot open every attached document are left out, and the approval card names them. Write it in the user's voice, first person, short, naming the hotel and what they need. Send only after the user agrees to send it.

team_message_send

Undo a combination: bring one listing that Anny AI combined into a tracked event back as its own tracked event. Anny AI never combines it again. Call it when the user says a combined listing is a different event.

tracked_event_separate

Show report email recipients, delivery status, invitation status, and existing teammates for a named hotel. Use hotels_list to resolve hotel names.

report_recipients_list

List who the user can send an App Message to: teammates (memberId), Hotel Access Groups (groupId) and how many Owners and Admins there are. With managedHotelId, only those who can open that hotel. With documentIds, only those who can open every document; leftOut names the teammates and groups who cannot, and which documents they cannot open.

team_message_recipients_list

List the user's App Message conversations with teammates, newest activity first: who is in each, its hotel, the latest message, who wrote it, and whether it is unread. Link a conversation as [the teammate's name](/app/messages/ followed by its id).

team_messages_list

Stop tracking one event for a hotel, along with the listings combined into it. Call it only when the user asks to stop tracking or remove that event.

tracked_event_untrack

Submit a product update or enhancement request to the Anny AI team on behalf of the connected user.

suggestion_submit