Plubox

Download one public HTTPS receipt image or PDF, at most 30 MB, and send it to the configured AI provider to identify purchased line items. Private hosts and redirects are rejected. Returns up to 500 items without storing a receipt or creating expenses; use a stored receipt workflow when a saved paginated classification is needed. Confirm sending the file to the AI provider before execution. Repeating the call sends it again and can incur another provider request. Category suggestions require review before financial entry. Use list-organizations, create-receipt or get-expense to identify the receipt. File IDs come from upload-receipt-files or the paginated get-receipt response. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

classify-receipt-image

Send the files of one authorized receipt to the configured AI provider and replace its saved classification with the resulting line items, using the website classification operation. This does not create expenses or payments. Results include the first 100 items; use get-receipt for remaining saved pages rather than repeating classification. Confirm sending these receipt contents to the AI provider first. Use list-organizations, create-receipt or get-expense to identify the receipt. File IDs come from upload-receipt-files or the paginated get-receipt response. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

classify-receipt

Commit the complete confirmed expense batch in one transaction after checking the exact revision, item count and digest. The received count must equal the immutable begin count. Current permissions, billing, receipt and every referenced type, gateway, accommodation and performer are rechecked. This creates the same expense, accepted payment, payroll and receipt effects as the website bulk form; it does not charge a card or call a payment provider. Each occurrence also receives a complete encrypted result snapshot; any write, snapshot or response failure rolls back the entire batch. Repeating this exact commit with the original live token returns the stored receipt without creating another expense, for up to 24 hours after commit or original token expiry, whichever comes first. Read all created IDs in order using get-expenses-bulk-results and a saved result using get-expenses-bulk-result with draft_id and ordinal. Individual snapshots expire within 24 hours of their capture or original token expiry; their deadlines can precede the batch receipt deadline. Use list-organizations, search-expense-references and search-accommodations for authorized IDs. A shared receipt_id must already exist in this organization. Use draft_id, revision and next_chunk_index returned by the draft tools; confirm the complete batch before commit. The record author always comes from the authenticated PMS actor. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

commit-expenses-bulk-draft

Create one task with the same field validation and events as the website. Optional photos use the documented MCP attachment transport. The complete result is captured atomically in an immutable result_receipt. Descriptions preview 1024 Unicode code points; photos use a bounded whole-item preview. Read all remaining content with get-create-task-result, using the receipt id and snapshot_hash. This continuation requires this exact write permission, not a task read permission, and never repeats the mutation. It is bound to the same live MCP token and expires after at most 24 hours or at the original token expiry. Input and result envelopes are limited to 1 MiB. Resolve tasks with list-tasks or list-task-table; types/rules with list-task-types/list-task-rules; targets with search-task-references or search-accommodations; photos with get-task and images_meta pagination. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

create-task

Creates one website-only long-term monthly rental price period. First call list-long-term-rental-periods and use an exact accommodation_group_id from available_groups. Periods for the same group may not overlap. The organization currency is applied automatically. This tool writes data and sends the change through the standard website synchronization channel.

create-long-term-rental-period

Create 1 to 50 expenses atomically with the website bulk expense contract. Every item requires cost_type, price, date, payer_type and payment.gateway_id; account-scoped users must also specify an accessible accommodation_id. A top-level receipt_id may attach one existing receipt to the batch. Any invalid item or result snapshot rejects the whole batch. Returns created IDs and a distinct immutable result_receipt for every zero-based input ordinal, including identical rows. Read each complete saved result with get-create-expenses-bulk-result-head, then follow get-create-expenses-bulk-result for text and receipt images under this same write permission. Each snapshot expires within 24 hours of capture or original token expiry, whichever is earlier; read its exact expires_at. This creates financial records and must never be automatically retried. Call list-organizations, search-expense-references and search-accommodations for authorized IDs; call list-expenses before editing an existing expense. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

create-expenses-bulk

Create a receipt container in the organization. It starts inactive and unclassified, as on the website. Upload files with upload-receipt-files, optionally classify them, then attach receipt_id when creating expenses. This creates a new container on every call. Use list-organizations, create-receipt or get-expense to identify the receipt. File IDs come from upload-receipt-files or the paginated get-receipt response. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

create-receipt

Create one expense with canonical website form fields, including payment.gateway_id, optional owner reimbursement timestamp and existing receipt_id. This uses the same validation, performer links, accepted payment and receipt activation as the website. The record author comes from the authenticated account. Input is limited to 512 KiB and the database TEXT capacity. Long response descriptions, notes and receipt images have bounded previews; use get-create-expense-result with result_receipt to read the complete immutable result under this same create permission. For direct chat attachments use add-expense. Confirm financial details before creation and never automatically retry. Call list-organizations, search-expense-references and search-accommodations for authorized IDs; call list-expenses before editing an existing expense. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

create-expense

Delete one accessible task and emit the same website deletion event. This removes the task; no retry creates a replacement. Resolve tasks with list-tasks or list-task-table; types/rules with list-task-types/list-task-rules; targets with search-task-references or search-accommodations; photos with get-task and images_meta pagination. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

delete-task

Delete selected task photos. Every photo must belong to an accessible task; mixed allowed/denied batches make no changes. Resolve tasks with list-tasks or list-task-table; types/rules with list-task-types/list-task-rules; targets with search-task-references or search-accommodations; photos with get-task and images_meta pagination. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

delete-task-photos

Soft-deletes one accessible Plubox expense and its related payment, using the same expense permissions as the website. Own access permits only expenses entered by the current user; full access permits expenses of other users within the authorized organization and account scope. Use an exact expense ID returned by add-expense or list-expenses. Never guess an ID. This destructive tool may only be called after the user explicitly asks to delete that expense.

delete-expense

Soft-deletes one website-only long-term monthly rental price period. First call list-long-term-rental-periods and use an exact period ID. Never guess an ID. Call this destructive tool only after the user explicitly requests deletion. The deletion is sent through the standard website synchronization channel.

delete-long-term-rental-period

Soft-delete the exact authorized expense, its payment and its payroll accrual using the same shared operation as the website. Requires the original delete permission, current organization/account and own-expense scope. A missing, foreign or already deleted record returns not_found. Returns the complete native deleted expense fields through bounded previews and an immutable result_receipt; use get-delete-expense-record-result for long description and performer notes, without repeating deletion. The payment and receipt follow the native post-deletion state. Confirm the expense before calling. Never automatically retry an uncertain delete. The original delete-expense tool retains its seven-field contract. Call list-organizations, search-expense-references and search-accommodations for authorized IDs; call list-expenses before editing an existing expense. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

delete-expense-record

Delete selected receipt images or PDF files within the current organization. Like the website deletion action, unavailable IDs are skipped. The response explicitly identifies deleted and skipped IDs from the request; it does not disclose another organization file. This does not delete the receipt or existing expenses. Confirm the exact files before deletion. Use list-organizations, create-receipt or get-expense to identify the receipt. File IDs come from upload-receipt-files or the paginated get-receipt response. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

delete-receipt-files

Read the optional installment schedule and current booking context. Required: organization_id and booking_id. Use the returned revision, currency, total and booking_state_hash for a confirmed schedule change. Access, validation, missing-record, billing and conflict errors are returned as structured errors. Never blindly retry a schedule write. list-organizations → list-bookings → get-booking-payment-schedule. Search accommodations before using an accommodation filter. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-booking-payment-schedule

Resolve the existing effective cleaning price for one accessible group and date. Find cleaning IDs with list-cleaning-records, get-cleaning-schedule or list-assigned-cleanings; Booking IDs with list-cleaning-bookings; group and setting IDs with list-group-cleaning-settings and list-cleaning-settings-for-group; employees with search-task-references. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-cleaning-price

Read enabled cleanings for confirmed Bookings on one date. Counts describe returned rows; meta.total describes all matches. Continue with meta. Find cleaning IDs with list-cleaning-records, get-cleaning-schedule or list-assigned-cleanings; Booking IDs with list-cleaning-bookings; group and setting IDs with list-group-cleaning-settings and list-cleaning-settings-for-group; employees with search-task-references. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-cleaning-schedule

Count enabled cleanings for confirmed Bookings on up to fifteen dates, using current assignment scope. Find cleaning IDs with list-cleaning-records, get-cleaning-schedule or list-assigned-cleanings; Booking IDs with list-cleaning-bookings; group and setting IDs with list-group-cleaning-settings and list-cleaning-settings-for-group; employees with search-task-references. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-cleaning-schedule-counts

Continue one fixed section of an immutable bulk expense creation result. Pass result_receipt.id and snapshot_hash, an exact section and offset=0 initially; follow meta.next_offset until has_more is false. Text pages contain at most 2048 Unicode code points and preserve null. Receipt image pages contain at most 50 complete images, reduced only by the response byte budget; next_offset follows the actual count. Works for both finite and staged bulk snapshots. Every call requires the original live PMS actor, organization and token, the exact bulk write permission, current billing access and current original-record scope. No general expense read permission is granted. Values come only from the immutable creation snapshot; later record changes are never mixed into it. Expiry is the saved result_receipt.expires_at, within 24 hours of capture or original token expiry. Reading never repeats financial writes. Use the original organization_id and result_receipt from create-expenses-bulk, or draft_id and zero-based ordinal from get-expenses-bulk-results. Never substitute a current expense ID for a result_id. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-create-expenses-bulk-result

Read one complete saved expense creation result using result_receipt.id and snapshot_hash from create-expenses-bulk. Long descriptions and performer notes have 1024-code-point previews; receipt images have whole-row previews. Continue every remaining section with get-create-expenses-bulk-result. Every call requires the original live PMS actor, organization and token, the exact bulk write permission, current billing access and current original-record scope. No general expense read permission is granted. Values come only from the immutable creation snapshot; later record changes are never mixed into it. Expiry is the saved result_receipt.expires_at, within 24 hours of capture or original token expiry. Reading never repeats financial writes. Use the original organization_id and result_receipt from create-expenses-bulk, or draft_id and zero-based ordinal from get-expenses-bulk-results. Never substitute a current expense ID for a result_id. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-create-expenses-bulk-result-head

Read this expense batch status, counts, revision, digest, next chunk index and expiry. Stored descriptions, payment input and receipt header are never returned. Expiry is the input deadline while open and the retained receipt deadline after commit. Original expense-write permissions and billing access are rechecked. Use list-organizations, search-expense-references and search-accommodations for authorized IDs. A shared receipt_id must already exist in this organization. Use draft_id, revision and next_chunk_index returned by the draft tools; confirm the complete batch before commit. The record author always comes from the authenticated PMS actor. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-expenses-bulk-draft

Read one saved creation result from a committed staged batch, identified by draft_id and its zero-based ordinal from get-expenses-bulk-results. Identical input rows retain distinct occurrences. Returns the bounded full DTO head and its result_receipt; follow get-create-expenses-bulk-result for remaining text or images. Batches created before result snapshots were introduced have no recoverable saved head and return unavailable. Every call requires the original live PMS actor, organization and token, the exact bulk write permission, current billing access and current original-record scope. No general expense read permission is granted. Values come only from the immutable creation snapshot; later record changes are never mixed into it. Expiry is the saved result_receipt.expires_at, within 24 hours of capture or original token expiry. Reading never repeats financial writes. Use the original organization_id and result_receipt from create-expenses-bulk, or draft_id and zero-based ordinal from get-expenses-bulk-results. Never substitute a current expense ID for a result_id. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-expenses-bulk-result

Read immutable creation IDs from a committed expense batch in original zero-based input order, up to 100 per page. Follow meta.next_offset until has_more is false. IDs remain in this historical receipt even if an expense is later deleted. Use get-expenses-bulk-result with draft_id and an ordinal for the complete saved creation result, under the same original write permission and current record scope. No descriptions or financial payloads are included in this ID page. The original actor, organization, live token, expense-write permissions and billing access are rechecked. IDs expire with the stored receipt, within 24 hours after commit or original token expiry; individual result snapshots have their own possibly earlier expiry. Older batches without stored snapshots retain this ID listing. Use list-organizations, search-expense-references and search-accommodations for authorized IDs. A shared receipt_id must already exist in this organization. Use draft_id, revision and next_chunk_index returned by the draft tools; confirm the complete batch before commit. The record author always comes from the authenticated PMS actor. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-expenses-bulk-results

Returns the Plubox daily schedule for confirmed bookings visible to the authenticated user. Includes check-ins, check-outs, current stays, guests, services, payment balances, tasks, parking information, and explicit gross and no-commission prices. price is the gross total including services and parking; price_without_commission is the corresponding total after source commission. booking_price and booking_price_without_commission isolate the stay price. Never subtract source_commission from either no-commission field again.

get-schedule

Returns one complete Plubox statistics block: overview, Pareto analysis, received bookings, or expenses. For block=bookings, exact start_date/end_date ranges take priority over months. Booking records are selected by booking.created_at without a current-status filter; full stays are not clipped. The result includes bookings, sold nights, ALOS, stay-length buckets, historical active-accommodation denominators, per-accommodation rates, channel shares, gross booking.price revenue/ADR, raw sources, and source_groups such as aggregated Booking.

get-statistics

Read the latest stored USD exchange-rate snapshot and its update time. This does not fetch new rates. Missing or invalid snapshots return unavailable; valid currencies are alphabetically paginated. Follow meta.next_page for all rates. These reads require an authenticated PMS connection with no organization selection. No user, owner, application or organization ID can be supplied. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-public-exchange-rates

Read one service and its confirmation, price editing and per-booking rules. Call list-organizations, list-service-types and list-services. Use only IDs returned for the intended organization. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-service

Read one system service type with its translations. Call list-organizations, list-service-types and list-services. Use only IDs returned for the intended organization. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-service-type

Read one exact Booking source in the selected organization, including its scoped rate-plan association and newest 25 commission entries. Follow source_commissions_meta with get-source-commissions for the rest. Use organization_id from list-pms-organizations and source_id from query-sources or list-sources. Resolve global_source_id with list-global-sources and reuse only an activated rate plan from the same organization. Never guess linked IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-source

Read the complete commission history of an organization Booking source in effective-date descending order, then ID descending order. To continue a source response, request page 2 with limit 25; follow meta.next_page thereafter. No commissions are changed. Use organization_id from list-pms-organizations and source_id from query-sources or list-sources. Resolve global_source_id with list-global-sources and reuse only an activated rate plan from the same organization. Never guess linked IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-source-commissions

Read a fixed task description through the exact tasks.list.typed access scope, without writing or loading its full TEXT value. Pages contain up to 2048 Unicode code points. Continue from description_meta.next_offset with expected_hash=description_meta.hash; concatenate description fragments in offset order. A changed value returns conflict even when updated_at is unchanged. Null remains null. Board content repeats the current scope, filters and employee assignment; target content requires the same related_to/related_id. A nested type must still match task_type_id and belong to the organization. Each call checks current membership and route permission. Use IDs and description_meta from the matching task, task-type or task-rule read. Preserve source board filters or typed target; never guess IDs or choose another scope to bypass denied access. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-target-task-content

Read one accessible task, target, completion metadata and photos. Continue images_offset using images_meta.next_offset. The form preserves complete task-type metadata, target floor and the existing building-only location; location is omitted when no scoped building address exists. A location content_meta entry identifies a raw preview: continue its fixed field through get-task-location-content using content_target and the same task ID. For php-array-json, concatenate raw fragments, decode to a PHP associative value, then apply Laravel Resource recursive numeric-key reindexing; all-numeric arrays are reindexed, mixed keys are retained and an empty object becomes an empty array. Depth only chooses inline versus raw representation, never limits the source value. Descriptions use a 1024 Unicode-code-point preview. When description_meta is present, continue with get-task-content using next_offset/hash and the same IDs, target and board filters. For a nested task type use field=type_description and its task_type_id. Content continuation is read-only and never repeats a write. Resolve tasks with list-tasks or list-task-table; types/rules with list-task-types/list-task-rules; targets with search-task-references or search-accommodations; photos with get-task and images_meta pagination. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task

Read a fixed task description through the exact tasks.list access scope, without writing or loading its full TEXT value. Pages contain up to 2048 Unicode code points. Continue from description_meta.next_offset with expected_hash=description_meta.hash; concatenate description fragments in offset order. A changed value returns conflict even when updated_at is unchanged. Null remains null. Board content repeats the current scope, filters and employee assignment; target content requires the same related_to/related_id. Each call checks current membership and route permission. Use IDs and description_meta from the matching task, task-type or task-rule read. Preserve source board filters or typed target; never guess IDs or choose another scope to bypass denied access. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task-board-content

Read a fixed task description through the exact tasks.read access scope, without writing or loading its full TEXT value. Pages contain up to 2048 Unicode code points. Continue from description_meta.next_offset with expected_hash=description_meta.hash; concatenate description fragments in offset order. A changed value returns conflict even when updated_at is unchanged. Null remains null. Board content repeats the current scope, filters and employee assignment; target content requires the same related_to/related_id. A nested type must still match task_type_id and belong to the organization. Each call checks current membership and route permission. Use IDs and description_meta from the matching task, task-type or task-rule read. Preserve source board filters or typed target; never guess IDs or choose another scope to bypass denied access. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task-content

Read one of four fixed fields from the building location exposed by get-task: address, location_data, location_translation_data or location_translation_description. Repeat content_target exactly; every call checks tasks.read, current membership, the task target and the same scoped building location and selected translation. There is no account-address fallback. Address reconstructs both title and address. Pages contain 2048 Unicode code points; concatenate raw content using next_offset/hash. php-array-json represents SQL JSON: decode the complete string associatively and apply Laravel Resource recursive reindexing of arrays whose keys are all numeric, preserving mixed keys; empty objects become empty arrays and null remains null. Arbitrary JSON keys and deep source values remain available without a business depth limit. Changed text returns conflict even without a timestamp change; changed or inaccessible source identity returns not_found. Use get-task and its target.location.content_target/content_meta. Never substitute another task, accommodation, location or translation ID. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task-location-content

Read one accessible task, target, completion metadata and photos. Continue images_offset using images_meta.next_offset. Descriptions use a 1024 Unicode-code-point preview. When description_meta is present, continue with get-task-row-content using next_offset/hash and the same IDs, target and board filters. For a nested task type use field=type_description and its task_type_id. Content continuation is read-only and never repeats a write. Resolve tasks with list-tasks or list-task-table; types/rules with list-task-types/list-task-rules; targets with search-task-references or search-accommodations; photos with get-task and images_meta pagination. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task-row

Read a fixed task description through the exact tasks.index.single access scope, without writing or loading its full TEXT value. Pages contain up to 2048 Unicode code points. Continue from description_meta.next_offset with expected_hash=description_meta.hash; concatenate description fragments in offset order. A changed value returns conflict even when updated_at is unchanged. Null remains null. Board content repeats the current scope, filters and employee assignment; target content requires the same related_to/related_id. A nested type must still match task_type_id and belong to the organization. Each call checks current membership and route permission. Use IDs and description_meta from the matching task, task-type or task-rule read. Preserve source board filters or typed target; never guess IDs or choose another scope to bypass denied access. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task-row-content

Read one of the recurrence rules and their task templates. Descriptions use a 1024 Unicode-code-point preview. When description_meta is present, continue with get-task-rule-content using next_offset/hash and the same IDs, target and board filters. For a nested task type use field=type_description and its task_type_id. Content continuation is read-only and never repeats a write. Resolve tasks with list-tasks or list-task-table; types/rules with list-task-types/list-task-rules; targets with search-task-references or search-accommodations; photos with get-task and images_meta pagination. Never guess IDs. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task-rule

Read a fixed task description through the exact task-rules.read access scope, without writing or loading its full TEXT value. Pages contain up to 2048 Unicode code points. Continue from description_meta.next_offset with expected_hash=description_meta.hash; concatenate description fragments in offset order. A changed value returns conflict even when updated_at is unchanged. Null remains null. Board content repeats the current scope, filters and employee assignment; target content requires the same related_to/related_id. Each call checks current membership and route permission. Use IDs and description_meta from the matching task, task-type or task-rule read. Preserve source board filters or typed target; never guess IDs or choose another scope to bypass denied access. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task-rule-content

Read a fixed task description through the exact task-rules.list access scope, without writing or loading its full TEXT value. Pages contain up to 2048 Unicode code points. Continue from description_meta.next_offset with expected_hash=description_meta.hash; concatenate description fragments in offset order. A changed value returns conflict even when updated_at is unchanged. Null remains null. Board content repeats the current scope, filters and employee assignment; target content requires the same related_to/related_id. Each call checks current membership and route permission. Use IDs and description_meta from the matching task, task-type or task-rule read. Preserve source board filters or typed target; never guess IDs or choose another scope to bypass denied access. Errors contain a stable code and field details. Never automatically retry when operation_may_have_succeeded is true.

get-task-rule-list-content