MCP App Store
by KeenEthics
EscapeTab icon

EscapeTab MCP Server for ChatGPT

By ELVIS HENRIQUE PEREIRA

Overview

EscapeTab helps restaurant operators inspect company overviews, tab sales, shifts, fiscal documents, loyalty activity, and daily-menu orders through read-only MCP tools with pagination and reconciliation details.

How to connect

  1. ChatGPT: Open the EscapeTab listing in the ChatGPT app directory and choose Connect. ChatGPT listing EscapeTab exposes 16 tools here.

Tools

export_loyalty_contacts

ChatGPT
Export loyalty program contacts with unmasked phone numbers. Returns personal data. The stored consent covers the loyalty program terms, not marketing; use the consent fields to filter. Every call is recorded in the company audit log.

get_business_overview

ChatGPT
Company overview for orientation: name, currency, first and last day with tab sales, first and last day with daily-menu orders, and feature availability. Call this first.

get_daily_menu_order

ChatGPT
Get a single daily-menu order by date and id. All fields other than lines and linesNextCursor repeat identically on every page. lines is paged by index: when linesNextCursor is present, pass it back as lines_cursor to fetch the remaining lines; it is absent once every remaining line has been returned. Line item and category names longer than 500 chars are truncated with a marker. The internal phoneKey linkage field is stripped; look up the customer's loyalty identity with the loyalty tools instead. shareToken is always stripped since it is a live capability credential accepted by the public order-cancellation endpoint, never a display field. Daily-menu orders come from an internal operational feature (shared-link daily menu). Their revenue is a separate stream from tabs — never merge or double-count the two.

get_daily_menu_orders_summary

ChatGPT
Summarize daily-menu order counts, revenue (paid vs. free), top ordered items, and loyalty points moved for a date range. orderCount includes every order in the range regardless of payment status; paidRevenueCents, bucket revenue, and topItems count only genuinely settled revenue and exclude fully refunded orders. Scans orders in a stable order, up to 1000 per call, across as many days as needed to fill that page. On any multi-page fetch, topItems is only a slice-local ranking (it can miss the true range-wide winner); use the itemPartials field, present on every page of that fetch including the last, to merge complete per-item totals across all pages and reconstruct the real ranking yourself -- pass nextOrdersCursor back as orders_cursor to fetch the next page. Top-level fields and topItems describe the company currency only; when the range also has orders in other currencies, their full breakdowns (including their own topItems) are listed separately in totalsByCurrency. loyaltyPointsEarned and loyaltyPointsSpent are summed across all orders regardless of currency, since points are not money. Daily-menu orders come from an internal operational feature (shared-link daily menu). Their revenue is a separate stream from tabs — never merge or double-count the two.

get_fiscal_summary

ChatGPT
Summarize fiscal document totals by status for a date range. Compare with get_sales_summary to reconcile sold vs fiscally issued totals. Only production-environment documents are counted; homologation-environment documents (test issuances, not real sales) are excluded so they never inflate reconciliation totals. A document whose cancellation was rejected by SEFAZ (cancellation_rejected) is legally still an authorized NFC-e, so it is counted in the authorized totals; authorized.cancellationRejectedCount reports how many of those authorized documents are in that state, for reconciliation visibility. Top-level fields describe the company currency only; when the range also has documents in other currencies, their full breakdowns are listed separately in totalsByCurrency so no total is silently mixed across currencies. Scans fiscal documents in a stable order, up to 5000 per call; when nextDocumentsCursor is non-null, the response covers only that slice of documents -- pass it back as documents_cursor to summarize the next slice.

get_loyalty_account

ChatGPT
Get a single loyalty account by phone number, with its balance and transactions (page size 50, most recent first). Returns the unmasked phone number, since this is a single-account admin lookup. Prefer full +E.164 format (e.g. +14155552671); a bare local number is only interpreted as Brazilian when the company's own default language is pt-BR, matching how that company's own customer-facing phone entry resolves ambiguous input. When transactionsNextCursor is non-null, pass it as transactions_cursor to fetch the next page of transactions.

get_loyalty_summary

ChatGPT
Summarize the loyalty program. With no date range, returns current account and balance totals, scanning accounts in a stable order up to 2000 per call. With a date range, returns net points earned and redeemed in that window, including earn/redeem transactions plus signed manual adjustments and reversals (account-merge records are excluded to avoid double-counting); this path scans up to 2000 accounts and up to 5000 matching transactions per call, and resumes mid-account on the next call if one account alone has more transactions than fit in the remaining budget. It also queries at most 500 accounts per call regardless of how many transactions they have, so a call can return nextAccountsCursor non-null even with low totals when many accounts in the range are quiet. When nextAccountsCursor is non-null, the response covers only that slice -- pass it back as accounts_cursor to aggregate the next slice.

get_sales_summary

ChatGPT
Summarize closed and cancelled tab sales for a date range and count all currently open tabs across shifts. The global open-tab snapshot is returned on the first page only and is not additive across pages. Revenue counts closed tabs only; daily-menu orders are a separate stream. Scans tab summary fields (no line-item detail) in a stable order, up to 5000 per call; when nextTabsCursor is non-null, the response covers only that slice of tabs -- pass it back as tabs_cursor to summarize the next slice. Top-level fields (revenueCents, averageTicketCents, paymentMethods, ...) describe the company currency only; when the range also has tabs in other currencies, their full breakdowns are listed separately in totalsByCurrency so no revenue is silently mixed across currencies. revenueCents is always exact; discountCents and serviceChargeCents reflect tab-level adjustments and may exclude per-item discount/surcharge detail for some tabs (see field descriptions).

get_shift

ChatGPT
Get a single shift by id, with financials computed from the tabs attributed to it (tab.shiftId), broken down by currency. A tab closed during this shift but cancelled after the shift closed is still attributed here as cancelled, matching the shift summary maintained by the tab-finalization trigger. closedCount/cancelledCount are exact rollup values included only on the first page; revenueCents/serviceChargeCents are computed by scanning the shift's tabs, up to 2000 per call -- pass nextTabsCursor back as tabs_cursor to scan the rest when it is non-null.

get_tab

ChatGPT
Get a full tab (comanda) sale record, including entries and computed totals. computedTotals uses trigger-maintained item discount/surcharge aggregates, so it stays bounded across entry pages; legacy tabs without those aggregates are composed only up to 440 entries. Item-level discount/surcharge amounts fold into the net subtotal rather than being broken out separately. entries itself is paginated at up to 200 entries per response, read directly off the entries subcollection: when entriesTruncated is true, pass the returned nextCursor as entries_cursor to fetch the remaining entries. Item names and notes are capped at 500 characters, variation groups are capped at five and adaptively reduced when an entry needs more space, and transfer histories are capped at 100 references per item plus a byte budget, with transferredOutTruncated marking omitted references, so every entry remains retrievable within the response budget. customerPhone, when present, is returned unmasked since this is a single-record admin lookup; loyaltyToken is always stripped since it is a live capability credential accepted by public loyalty endpoints, never a display field.

get_top_items

ChatGPT
Return the best-selling items from closed tabs for a date range, with quantity and revenue in integer cents. Scans tabs in a stable order, up to 2000 per call, and hydrates their line items up to a bounded number of entry documents per call -- a single item-heavy tab can itself span multiple pages, resuming mid-tab. On any multi-page fetch, items is only a slice-local ranking (it can miss the true range-wide winner); use the itemPartials field, present on every page of that fetch including the last, to merge complete per-item totals across all pages and reconstruct the real ranking yourself -- pass nextTabsCursor back as tabs_cursor to fetch the next page. If the range spans more than one currency and currency is not set, the call is rejected as a tool error listing the currencies present -- retry passing currency to aggregate only tabs of that one. When currency is set, tabs of other currencies in the range are silently excluded from items/itemPartials (still scanned for pagination). If the range spans more than one currency and currency is not given, the first page establishes it for the whole fetch; a later page landing on a different currency is rejected as a tool error.

list_daily_menu_orders

ChatGPT
List compact daily-menu order records for a date range. Daily-menu orders come from an internal operational feature (shared-link daily menu). Their revenue is a separate stream from tabs — never merge or double-count the two.

list_fiscal_documents

ChatGPT
List fiscal document (NFC-e) records for a date range, optionally filtered by status. By default, homologation-environment documents (test issuances, not real sales) are excluded; pass environment: "homologation" to see them instead.

list_loyalty_transactions

ChatGPT
List loyalty point transactions for a date range, with masked phone numbers. Pagination is by customer account: each call scans up to 25 accounts and returns up to 200 transactions. If an account has more matching transactions than fit in the page, the cursor resumes mid-account on the next call so nothing is skipped or duplicated.

list_shifts

ChatGPT
List shift (turno) records for a business-day range, with opened/closed staff and per-shift sales summary.

list_tabs

ChatGPT
List compact tab (comanda) sale records for a business-day range. When status is open, the range is ignored so all open tabs, including those inherited from earlier shifts, are included. A cancelled row carrying a mergedInto field is a merge artifact, not a real cancellation: its items were merged into the tab identified by mergedInto. Sales summary tools already exclude these merge artifacts from cancellation counts.

Capabilities

Writes

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/escapetab"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/escapetab)