- Brand
- Finaloop
- Category
- Finance
- Primary Subcategory
- Accounting & Bookkeeping
Integration details
Description
Connect your AI assistant to Finaloop to review your financial reports, ledger and business performance.
- Integration type
- Plugin
- Verification status
- Not applicable
- Platform
- ChatGPT
- Primary Subcategory
- Accounting & Bookkeeping
- Secondary Subcategories
- None listed
- Brand
- Finaloop
- Access
- Account required
- First tracked
- 2026-11-22
- Tool count
- 51
- Market
- US
The Primary Subcategory used for this profile’s headline score.
Other Subcategories where the Integration is listed.
Not discoverable in ChatGPT yet
ChatGPT now suggests Plugins on its own when they match a user's request, but not every Plugin is being surfaced yet.We haven’t seen Finaloop suggested yet. Sign up and we’ll track it, and your Discoverability Score will appear here once ChatGPT starts surfacing it.
What discovery looks like

Get alerts for Finaloop
Get updates when Finaloop’s Discoverability Score or category rank changes.
Competing in ChatGPT Accounting & Bookkeeping
View Category51 tools agents can invoke
Compares what was paid to Meta, Google and Amazon with the ad spend counted from their own reports, month by month, over the question's months stretched to at least the last 3 full months. Call it with any ad spend or ROAS answer. Returns each platform's counted spend, payments, gap and whether it is flagged, plus total ad spend, true ROAS and new-customer ROAS with and without the payments of flagged platforms. A platform is flagged when, over the window (the question's months stretched to at least the last 3 full months), payments to it ran 25%+ above its own report and the gap is over $5,000 per month of the window. For each flagged platform give the gap and the window, say it may be ad accounts not connected to Finaloop (or a connection that started part-way), and give total_ad_spend_with_payments, true_blended_roas_with_payments and true_blended_nroas_with_payments next to the usual figures; added_to_spend is what the payments add to the spend already counted (months with no report already count their payments). Spend itself stays the platform's report. A smaller gap is billing timing: say nothing about it.
check_ad_payments_against_reports
Splits Amazon ad spend, sales, true ROAS and new-customer ROAS between FBA and FBM for whole months, all Amazon marketplaces together. Each advertised product's spend follows its own Amazon sales by channel, else its merged product's; the rest of the month's Amazon ad spend follows the month's FBA / FBM sales. Offer it when the user asks about Amazon ROAS and the company sells through both channels; run it only when they want it. Give ad spend, true ROAS and new-customer ROAS per channel, and say: the spend is allocated by product (each advertised product follows its own Amazon sales by channel, else its merged product's; the rest follows the month's FBA / FBM sales); FBM's new-customer ROAS is entirely estimated from FBA's known buyers (give amazon_new_customer_coverage; at 0 say new-customer ROAS cannot be estimated, never 0); a low ad_spend_matched_by_product_share (before an Amazon account's report cutoff) means both channels read alike; show unallocated_ad_spend when not zero (months with Amazon ad spend and no Amazon sales). When credits exceed the spend with no product, that negative rest follows the month's sales mix, so a channel can show a small negative spend or ROAS: say so rather than hide it. It covers Amazon ads and sales, all marketplaces together.
get_amazon_roas_by_fulfillment_channel
Status of a background export started by export_ledger or export_general_ledger (QUEUED, RUNNING, READY, FAILED, or CANCELLED, with row progress). When READY the response also carries downloadUrl, a signed link valid for expiresInSeconds (600) — after expiry, calling this tool again returns a fresh link.
get_report_export_status
Business-analysis tool: aggregations, trends and breakdowns over GMV, ad spend and unit economics — answers that are numbers, computed on an order-level GMV basis rather than booked P&L accounting. For individual records — orders, bills, invoices, bank transactions, purchase orders — and workflows that change them, the GraphQL tools fit better: they return each record in its natural shape, with display numbers, customers, totals and statuses. Executes a read-only query in FDQL (Data Query Language). FDQL is a JSON-based query language, not SQL; the full syntax reference is available via the business_analytics_get_fdql_syntax tool. The query parameter is FDQL JSON. Example: {"version":"1.0","dataset":"companies","select":[{"alias":"name","expression":"@name"}],"limit":10} Results return at most 1000 rows (`truncated: true` when more exist) — aggregate, filter, or page with limit + offset for larger sets.
business_analytics_execute_query
Business-analysis tool: aggregations, trends and breakdowns over GMV, ad spend and unit economics — answers that are numbers, computed on an order-level GMV basis rather than booked P&L accounting. For individual records — orders, bills, invoices, bank transactions, purchase orders — and workflows that change them, the GraphQL tools fit better: they return each record in its natural shape, with display numbers, customers, totals and statuses. Returns distinct values for a field in an FDQL (Data Query Language) dataset, with occurrence counts where the dataset's backing store supports counting (values come back without counts otherwise). The FDQL syntax reference is available via the business_analytics_get_fdql_syntax tool. A field given as "@facet:<id>" instead returns that facet's values for the company as {id, label, count} — filter with the id via {"@facet:<id>": {"$in": [...]}}. There, count is a raw row count over the underlying dataset used for ranking values by activity; it is not a business metric (not orders, not sales).
business_analytics_get_distinct_values
Business-analysis tool: aggregations, trends and breakdowns over GMV, ad spend and unit economics — answers that are numbers, computed on an order-level GMV basis rather than booked P&L accounting. For individual records — orders, bills, invoices, bank transactions, purchase orders — and workflows that change them, the GraphQL tools fit better: they return each record in its natural shape, with display numbers, customers, totals and statuses. Returns the schema for one FDQL dataset: field names, types, relationships, and any documented formulas. Available datasets: accountsMetadata, adAccountBillingCharge, adAccount, adCampaign, adSellerPayment, adSet, affiliatedEntities, affiliatedEntitiesCompanyOverrides, affiliatedEntityGroupsMappings, bankAccountOverview, bankAccountDuplicationGroups, bankAccountPrimaryMappings, bankAccounts, bankTransactions, bills, chargeback, companies, costAdjustmentItem, customerAddresses, customerCohorts, customers, entityMetrics, entityMetricsSpendTypes, holders, institutions, integrationSalesChannels, inventoryPurchase, operationalAccounts, payrolls, peerBenchmarkDefinitions, peerCohortMetricsV2, peerCohortsV2, peerVendorCostBenchmarks, platformRoles, productMerges, products, stores, udfRunningBalance, vendors, allSales, sourceDescriptions, contributionMargin, allCashSpend, adPerformance, marketingEfficiency, merchants, primaryBankAccounts, canonicalStores. FDQL is a JSON-based query language, not SQL; its full grammar is available via the business_analytics_get_fdql_syntax tool.
business_analytics_get_dataset_schema
Business-analysis tool: aggregations, trends and breakdowns over GMV, ad spend and unit economics — answers that are numbers, computed on an order-level GMV basis rather than booked P&L accounting. For individual records — orders, bills, invoices, bank transactions, purchase orders — and workflows that change them, the GraphQL tools fit better: they return each record in its natural shape, with display numbers, customers, totals and statuses. Returns the complete FDQL (Data Query Language) reference: syntax, operators, functions, filter and join grammar, and worked examples. FDQL is a JSON-based query language, not SQL.
business_analytics_get_fdql_syntax
Business-analysis tool: aggregations, trends and breakdowns over GMV, ad spend and unit economics — answers that are numbers, computed on an order-level GMV basis rather than booked P&L accounting. For individual records — orders, bills, invoices, bank transactions, purchase orders — and workflows that change them, the GraphQL tools fit better: they return each record in its natural shape, with display numbers, customers, totals and statuses. Returns sample rows from an FDQL (Data Query Language) dataset, optionally projected to specific fields. The FDQL syntax reference is available via the business_analytics_get_fdql_syntax tool.
business_analytics_get_sample_rows
Business-analysis tool: aggregations, trends and breakdowns over GMV, ad spend and unit economics — answers that are numbers, computed on an order-level GMV basis rather than booked P&L accounting. For individual records — orders, bills, invoices, bank transactions, purchase orders — and workflows that change them, the GraphQL tools fit better: they return each record in its natural shape, with display numbers, customers, totals and statuses. Lists the FDQL (Data Query Language) datasets available to the caller (name, displayName, type, description). Some datasets also support published facets — richer, cross-dataset groupings referenced via "@facet:<id>"; call business_analytics_get_dataset_schema on a specific dataset to see which ones apply.
business_analytics_list_datasets
The quick fix for a bill with the wrong or missing categories: replaces its categorized amounts in one call without touching the vendor, dates, memo or bill number. Each category is tagged with a classification category by name (e.g. "software subscriptions") or id, and the bill total is recomputed from the new set. For changes beyond the categories, use update_bill.
categorize_bill
The quick fix for an invoice with the wrong or missing category: retags every non-shipping line with one classification category in a single call, without touching the customer, dates, line items, memo or invoice number. For changes beyond the category, use update_invoice.
categorize_invoice
Completes a step of the year-end Orders checklist on the person's behalf, for the year-end under review: `balances` confirms the unpaid-orders balance shown by get_orders_year_end (stored as that year's closing balances so a later drift can be detected), `submit` closes the Orders domain once every step is done. The answer names the year. Refused when no year-end is under review, when the review changes while the balances are read, or when the step cannot move to that state.
complete_orders_year_end_task
Creates a bill in one call: resolves the vendor by id or name (creating it when the name is new), resolves each category line to a classification tag by name or id, and saves the bill. A bill is described here as who it is from, when it was dated and when it is due, and what it was for as one or more categorized amounts. Retries can pass the billId returned by a previous attempt — the call is idempotent per id. Linking the bill to the bank transaction that paid it is match_bill_payment; deleting a bill happens in the Finaloop app for now. To edit an existing bill use update_bill, or categorize_bill to just fix its categories. The parameter names here belong to this tool rather than to the person asking: a bill is described as a vendor, dates, and some categorized amounts, so gathering those details in conversation fits better than walking someone through this schema. preview_bill shows this call's resolved result without saving; when the vendor or categories have not been confirmed in conversation, previewing first and presenting the outcome avoids surprises like an unintended new vendor.
create_bill
Creates an invoice in one call: resolves the customer by id or name (creating one when customerEmail is also given), resolves each line to a product by SKU when given, and saves the invoice. A SKU that matches no product is an error, not an auto-create; omit sku for a free-form line and give description and amount instead. By default the invoice is saved as a draft; passing publish: true issues it and requires a due date and a category on every line. Retries can pass the invoiceId returned by a previous attempt — the call is idempotent per id. Line categories take a few moments to settle after this call returns, so a follow-up edit made right away can still see the invoice as uncategorized. Paying, voiding and deleting an invoice happen in the Finaloop app for now. To edit an existing invoice use update_invoice, or categorize_invoice to just fix its categories. The parameter names here belong to this tool rather than to the person asking: an invoice is described as a customer, some line items and dates, so gathering those details in conversation fits better than walking someone through this schema.
create_invoice
Creates a purchase order in one call: resolves the vendor by id or name (creating it when the name is new), resolves each line to a product by SKU (creating missing products), and saves the PO. A SKU that doesn't exactly match one of the company's products (after trimming surrounding spaces; case and dashes matter) creates a new whole-unit product; that product stays even if the purchase order is later deleted or the save fails, and products can't be deleted. Looking the SKU up among the company's existing products first and confirming with the person before sending a SKU that matches none avoids that; a near-identical existing SKU usually means a typo. Purchase orders exist only on the unit tracking (Inventory IQ) plan — for a company on another inventory plan the call returns an UNSUPPORTED_MANAGEMENT_METHOD error whose message links to the Inventory IQ page in the Finaloop app. By default the PO is saved as a draft for review in the Finaloop app. Passing publish: true moves it to in-progress and requires placementDate, inProgressDate and a directCostPerUnit on every line. Retries can pass the purchaseOrderId returned by a previous attempt: while that order is still an unpublished draft with no shipment groups, the retry rewrites it rather than creating a second order. Any other existing id — a published, deleted or demand-planning order, one with shipment groups, or another company's — comes back as PURCHASE_ORDER_ALREADY_EXISTS; after a retried publish that usually means the first attempt went through, so reading the order beats creating it again. Quantities are whole numbers unless the SKU is an existing fractional product (SKUs created here are whole-unit products), inProgressDate falls on or after placementDate, and a published order takes no date after the end of the current year. A rule the saved order breaks comes back as INVALID_INPUT whose message starts with the rule's code; an inProgressDate inside a closed-books (locked) period is rejected that way with LOCKED_PERIOD. To edit an existing purchase order use update_purchase_order instead. The parameter names here belong to this tool rather than to the person asking: an order is described as a vendor, some products, quantities and prices, so gathering those details in conversation fits better than walking someone through this schema.
create_purchase_order
Returns a company's customer lifetime value exactly as its Lifetime value dashboard shows it: customers are grouped into cohorts by the month of their first direct-to-consumer or marketplace order, and each cohort is followed month by month (M0 is the acquisition month). The response holds the six headline cards (first-order profitability, True LTV, True LTV:CAC, customer contribution, CAC payback in months, repurchase rate) with their values for the equal-length period before and the change, plus one table per requested metric: a row per cohort month with new customers, CAC, first order, M0..Mn and overall, and Total (additive metrics) and Average rows. Bases: sales are order-level (GMV); gross profit is after product cost, selling, fulfilment and processing fees and estimated refunds for months still settling; CAC is ad spend allocated to stores plus affiliate fees and product giveaways; amounts are in the company home currency; only finished months are included. A card with isFloor (previousIsFloor for the prior period) is a lower bound: CAC payback shown as 3 with isFloor reads as 3 months or more. With no acquisition cost recorded, CAC-based ratios are left empty while contribution and payback count it as zero. Figures are measured unless predictive is on. With predictive, the months a cohort has not reached yet are forecast from how older cohorts developed and from seasonal patterns, as the dashboard's Predictive mode does; each table then carries a forecast section naming, per cohort, the first forecast month and the likely error of each forecast cell in percent (given for the cumulative metrics only), and how many cohorts in each Average and Total cell are forecast. On a forecast row, overall takes in the forecast months, so it is a forecast too. For a cumulative metric, a forecast cell with no error figure lies further out than the error was ever measured. The headline cards and RETENTION_RATE are always measured. With compare_by and compare_items the response is the dashboard's Compare view instead: per metric, one row per compared store or SKU holding that item's typical cohort (its Average row) over the period. Each item is its own set of customers, acquired with that store or SKU in the first order; a store is charged its own acquisition cost, while a SKU is charged the company's acquisition cost as a share of first-order sales, applied to that SKU's first orders, since ad spend is not tracked per SKU. One comparison covers either stores or SKUs. A comparison can be narrowed by the other axis: skus when comparing stores keeps each store's customers whose first order there included those SKUs, and each store is then charged those SKUs' share of its own acquisition cost; stores when comparing SKUs keeps each SKU's customers who first bought it at those stores, and the SKU's share is of those stores' acquisition cost. Metrics: SALES_PER_CUSTOMER, GROSS_LTV (cumulative sales per customer), TOTAL_SALES, CUMULATIVE_ORDERS, AOV, GROSS_PROFIT_PER_CUSTOMER, TRUE_LTV (cumulative gross profit per customer), TOTAL_GROSS_PROFIT, GROSS_PROFIT_RATE, RETENTION_RATE (share active in the month), NET_REVENUE_RETENTION (month sales against first-purchase sales), REPURCHASE_RATE (share who ever ordered again), LTV_TO_CAC, TRUE_LTV_TO_CAC, ROI, CONTRIBUTION_PROFIT (gross profit less CAC), PAYBACK (contribution per customer), REFUND_RATE, DISCOUNT_RATE. Store and SKU filters narrow who is in a cohort by what the first order was; after that all of the cohort spend counts. With a SKU filter, CAC is the company or selected stores' marketing as a share of first-order sales, applied to the SKU's first orders, so a cheaper SKU's customers cost less to acquire. Their values come from include_filter_options. Slicing cohorts by something this tool does not offer (country, product line, customer source) is a question for the FDQL query tools, on sales-based figures.
get_customer_lifetime_value
Runs a GraphQL query or mutation against this server's schema and returns the GraphQL response (data and errors). The operation goes through the same schema check `validate` runs, and its variables are coerced against their declared types, so only operations this server exposes are accepted; subscriptions are not supported. Reading the schema is what `introspect` and `search` are for — on company-scoped servers a raw introspection query (`__schema`, `__type`) is rejected.
execute
Explains why a company's cost of goods sold for a month or range of months looks the way it does. It checks the period against the previous 12 months and the activity behind it, and returns a verdict with findings: information that is incomplete and can be completed in the app (each with a link), or reasons the number legitimately differs from other months, such as shipping timing, returns, or a change in how cost of goods sold is calculated. Amounts are in the company's home currency and on the company's accounting method (accountingMethod: cash or accrual), matching its Profit & Loss in the app. The verdict is one of ACTION_NEEDED, EXPLAINED, NOTHING_UNUSUAL, or a short note that the analysis is still running and when to check back. Each cost line row compares this period's amount (periodAmount) with the average of the previous 12 months scaled to the same length (trailingAverageAmount); changePct is the percent difference. Amounts are positive costs in the home currency. netSales reports the period's net sales the same way: amounts in the home currency, with the trailing average taken over the previous 12 months and scaled to the same length.
explain_cogs
Starts a background CSV export of the full (unpaginated) ledger slice for a report line — same runId and ref as get_ledger. Returns a jobId to poll with get_report_export_status. A company can run at most 2 background exports at once; a TOO_MANY_REQUESTS error means one must finish first.
export_ledger
Exports a full report as a styled XLSX workbook, synchronously. Takes the same inputs as run_report and returns { url, filename }; url is a signed download link valid for 10 minutes. Very large reports fail with REPORT_TOO_LARGE — a narrower date range succeeds. Figures are on the company's accounting method.
export_report
Starts a background export of the company's full general ledger for a date range as an XLSX workbook. Returns a jobId. For most companies poll it with get_report_export_status (also listed by list_report_exports) to get the download link; a company still on the previous reporting system instead receives the workbook by email at the caller's address (deliveredByEmail: true) and the jobId cannot be polled. A company can have one general-ledger export in progress at a time; a TOO_MANY_REQUESTS error means it must finish first. For a company in list_my_companies reportsMode 'both', a period ending before 2026-01-01 is history kept on the previous reporting system and is exported (and emailed) from there, and a period crossing that date takes two calls, one per system; a reportsMinDate later than that date means the earlier history is locked or before the books. On the previous reporting system a range across a change of accounting method takes one call per method period. The response's accountingMethod is the method the ledger is on: the company's own method, which for a period on the previous reporting system is the one its books use for the fiscal year the period ends in.
export_general_ledger
This tool serves balance sheets as of dates before 2026-01-01; from that date on, use run_report. Fetch the balance sheet report for a company as of a date, as shown on the Finaloop web app. Returns the report tree as JSON. Figures are on the accounting method the company's books use at that date; the response states it.
get_balance_sheet
This tool serves periods ending before 2026-01-01; from that date on, use run_report. Fetch the cash flow statement for a company over a date range, as shown on the Finaloop web app. Returns the report tree as JSON. Defaults to totals only; pass totalsOnly=false and breakdowns for detail rows.
get_cash_flow
This tool serves periods ending before 2026-01-01; from that date on, use run_report. Fetch the individual ledger transactions behind a report line for a company and date range. Useful for drilling into a number seen on the profit & loss, balance sheet or cash flow report. Paginated, 100 rows per page. For a full-period dump do not page through this tool: for one account run the `ledgerAccountTransactionsCSV` query via the `execute` tool with its accountName, then poll `checkExportStatus` with the returned requestId for the download link; for every account export the full general ledger instead. Transactions are on the accounting method the company's books use at endDate; for a range that crosses a change of method the response covers the part on that method and names the rest.
get_ledger_account_transactions
The year-end checklist's Orders step, for the year-end currently under review (the answer names the year): the unpaid Shopify-orders balance, unlinked store cash payments and their net as of that fiscal year end, the step's state (open, in progress, closed), whether the balances were already confirmed, and the store connections still waiting for access. Errors when the company has no year-end under review.
get_orders_year_end
This tool serves periods ending before 2026-01-01; from that date on, use run_report. Fetch the profit & loss (income statement) report for a company over a date range, as shown on the Finaloop web app. Returns the report tree as JSON. Defaults to totals only; pass totalsOnly=false and a breakdown for per-category/vendor detail rows. Figures are on the company's own accounting method and the response states it; a range that crosses a change of the company's accounting method is answered with the separate per-method calls to make instead.
get_profit_and_loss
Returns the SDL definition of one named GraphQL type in this server's schema, with the types it references expanded `depth` levels deep. `Query` and `Mutation` list every available operation. `search` finds types by keyword when the exact name is not known.
introspect
The ledger drill-down for one report line: the individual entries behind a run_report figure. Takes the runId and one ref from run_report or expand_report_rows — rowRef for a row (a HEADER row's ref covers the whole section), cellRef for a horizontal-breakdown cell. Returns dynamic columns plus rows of positional cells, a sliceTotal for the whole slice, and nextCursor for pagination (pass it back as cursor for the next page). Entries can be filtered to particular values of a column — ask for that column's filterValues, then pass them as selections: [{ columnId, valueIds }]. Runs expire 12 hours after run_report.
get_ledger
Lists the companies the authenticated caller can access (id, name, accountingMethod, booksStartTime, reportsMode, reportsMinDate). The returned ids are the valid tenantId values for other tools. reportsMode says which report tools serve the company: 'legacy' (get_profit_and_loss, get_balance_sheet, get_cash_flow, get_ledger_account_transactions), 'new' (run_report and its companions), 'both' (either, split by date), or 'none'. reportsMinDate, when set, is the earliest period start run_report and export_report accept for that company; earlier periods belong to the previous-system report tools when those are listed. access is per company: 'full', 'read-only' (reads only; nothing can be created or changed there), 'basic' (your role there allows only inventory, bills and invoices) or 'inventory' (inventory only); the reportsMode of a 'basic' or 'inventory' company is 'none'.
list_my_companies
Lists the financial reports available to a company on Finaloop's reporting system: the default Profit & Loss, Balance Sheet and Cash Flow plus any custom reports. Returns ids usable as run_report's blueprintId. Each report lists the breakdowns it can be split by (isTime marks period breakdowns, parameters their options) and supportsAllocations, true when the report covers a period and offers at least one non-time breakdown, so allocations can be applied to it.
list_reports
A company's background exports, newest first (50 per page) — the recovery surface when a jobId is no longer at hand: each row carries the id (usable as get_report_export_status's jobId), status, type, file name, and row progress. Mirrors the app's Exports history page, so it lists the report center's own exports unless `domains` asks for another product's.
list_report_exports
Every published article of the Finaloop help center (help.finaloop.com), whose published articles explain how Finaloop and Margins work: setup, integrations, transactions, inventory and COGS, reports, year-end and tax, account management and release notes, as id and title. Useful for browsing when a search finds nothing.
list_help_articles
Links a bank transaction to the bills it paid, settling one bill or several in a single call, and relieving each bill's payable by the amount allocated to it. Bills belonging to different vendors can share one payment — each keeps its own payee. Allocating less than a bill's balance leaves the rest open as a partial payment. The bills plus an optional fee account for the whole payment; anything left over is reported instead of written. A category already covering the whole transaction is replaced by the match, the same as re-categorizing it in the Finaloop app; a partial split into other categories stays, with the bills covering the rest. Re-running it with a different set of bills replaces the previous match, and unmatch_bill_payment removes bills from one. preview_bill_payment shows the same split without writing it.
match_bill_payment
Returns a link to a page in the Finaloop web app. Use when the user asks to see, open, review, or go to something in the app, or when a task needs a surface these tools do not cover (attachments, shipments, connecting integrations). Pages: - transactions: Bank and credit-card transactions list. - transactions_requiring_input: Transactions waiting for the user to categorize or assign a vendor, grouped by vendor. - transaction: One bank transaction, opened in the transactions list. - unidentified_bank_transactions: Unidentified bank transactions, grouped by what they need: a payout or transfer to identify, or access to grant to a virtual bank such as Divvy, Ramp or PayPal. - bills: Bills list. - bill: One bill. - invoices: Invoices list. - invoice: One invoice. - invoice_payment_group: One customer payment covering one or more invoices. - orders_branded_stores: Orders from branded stores (Shopify and similar). - orders_marketplaces: Orders from marketplaces (Amazon, TikTok Shop and similar). - order_branded_store: One branded-store order. Use for Shopify and other own-store channels. - order_marketplace: One marketplace order. Use for Amazon, TikTok Shop and other marketplace channels. - customers: Customers list with sales and balances. - customer: One customer. - gift_cards: Shopify gift cards issued and redeemed. - orders_fulfillment_settings: Per-store setting for whether sales and cost of goods sold are recorded when orders are placed or when they ship. - inventory_products: Products list with stock and cost. - product: One product, opened in the products list. - inventory_cogs: Cost of goods sold overview. - inventory_purchases: Inventory purchases with their received dates. - purchase_orders: Purchase orders and their shipments. - vendors: All vendors the business spends money with. - payrolls_employees: Employee payroll runs. - payrolls_contractors: Contractor payments. - ads: Advertising spend across connected ad platforms. - banks: Bank and credit-card accounts overview. - loans: Loans and financing. - owners: Owner draws and contributions. - dashboards: Business dashboards home. - report_center: Report center — all reports and reconciliations. - report: One report in the report center, by report id. - balance_sheet: Balance sheet. - profit_and_loss: Profit and loss statement. - cash_flow: Cash flow statement. - journal_entries: Manual journal entries. - accruals: Accruals and prepaid expenses. - year_end: Year-end close checklist. - settings_integrations: Connect a new integration — sales channel, payment processor, bank, accounting tool, ads platform or loan provider. Pass appId to pre-select one. - settings_banks: Bank connection settings. - settings_billing: Finaloop subscription and billing. - launchpad: Setup checklist and onboarding tasks. - file_upload: Upload page for one upload session, by session id: the user uploads its files there.
open_app_page
Shows a report the way the allocation page does, for one saved allocation or a new one split by one or two breakdowns: the rows a rule can be set on, with the unallocated amount and current rule of each, the members money can be split to, the metrics, windows and fallbacks a rule can use with options.defaults (the window and fallback the web app picks when the user names none), and the allocation's periods. valuesDividedBy is the divisor already applied to every value and unallocated amount. Returns a runId that preview_report_allocation and save_report_allocation take; rows deeper than maxDepth are reached with expand_report_rows on the same runId. Runs expire after 12 hours.
open_report_allocation
Shows what create_bill would do without saving anything: which vendor the bill would go to — an existing one, or a new one that would be created — how each amount maps to a category, the computed total, and anything that would block creation such as a locked accounting period or an unknown category. Near-match vendor and category candidates come back too, so "did you mean …?" can be answered without further lookups. Suited to presenting the bill back to the person for confirmation or correction before creating it, especially when the vendor or a category was inferred rather than stated explicitly.
preview_bill
Shows what match_bill_payment would do without writing anything: how much of the payment lands on each bill, each bill's balance due and payee, and the allocation problems that would stop the match — amounts that do not add up, a bill allocated past its balance, currencies that cannot share a match. Conditions the classification write itself enforces, such as a locked period, surface when the match runs. A payment is consumed in full by the bills it pays plus an optional payment fee, so a set of bills that leaves part of the transaction unaccounted for comes back as a blocker rather than a partial match. A transaction that already carries a category for its full amount — as one a rule categorized on import does — is re-categorized as the bill payment, and the categories it replaces are listed under categoriesReplaced; a split that leaves part of the transaction uncategorized is kept, and the bills cover that part. Suited to presenting the split back to the person before committing it, especially when the amounts were inferred rather than stated.
preview_bill_payment
Shows the report from open_report_allocation as it would look with the given rule and period changes, without saving anything. Returns the slim grid with the allocated values, effects (per changed rule: the amount placed, the amount that could not be placed and why), notes, and blockers: reasons the setup would be refused when saved. valuesDividedBy is the divisor already applied to every value. skippedGroups lists the groups a split did not apply to because they lack a member it names. defaultsUsed lists the metric rules whose left-out window or fallback took the default, with the default each took. Takes changes only: rules and periods not mentioned keep their saved state. Preview cells can't be drilled into.
preview_report_allocation
One published article of the Finaloop help center: its title, description, public url to cite, last update time, and body as HTML trimmed to headings, paragraphs, lists, links and tables.
get_help_article
Returns the rows under one row of a report run (a run_report response, for one), in the same slim shape and with the same refs, without re-running the report: the rows that response omitted as collapsed. maxDepth counts from the given row. Runs expire 12 hours after they were made.
expand_report_rows
Runs a financial report and returns a slim grid plus a runId: per table its columns and a row tree with label, depth, kind and one value per column. Rows that can be drilled into or expanded carry a ref (r12); cells with their own drill-down carry cellRefs (r12:b). get_ledger, export_ledger and expand_report_rows take the runId with one ref. Rows deeper than maxDepth are omitted and their parent carries collapsed: true with childCount; expand_report_rows returns them from the stored run without re-running. valuesDividedBy is the divisor already applied to every value (1 when the values are in whole currency units, 1000 when the report displays thousands). A run stays available for 12 hours. Balance Sheet is as-of endDate (no startDate); Profit & Loss and Cash Flow cover startDate..endDate. Figures are on the company's accounting method; runtimeFilters in the response shows the method applied.
run_report
Saves changes to a report allocation, or creates one, from a runId of open_report_allocation: rules set on rows (a share of a metric over a time window, or a custom percent split across members), rows reset to unallocated, and periods added, re-dated or removed. Rules and periods not mentioned are kept as they are. With publish, the allocation is applied, so every report showing it uses the new rules; an allocation that is already applied is always saved this way, as in the web app. defaultsUsed lists the metric rules whose left-out window or fallback took the default, with the default each took. preview_report_allocation shows the result of the same changes without saving.
save_report_allocation
Full-text search over this server's GraphQL schema — type names, field names and their descriptions. Returns the SDL of each matching type or field together with its argument input types and return type, enough to write an operation. `introspect` returns one type by exact name.
search
Searches the Finaloop help center (help.finaloop.com), whose published articles explain how Finaloop and Margins work: setup, integrations, transactions, inventory and COGS, reports, year-end and tax, account management and release notes. Returns up to 10 matches ranked by keyword relevance, each with its id, title, description, public url and the matching excerpts; get_help_article returns an article's full text.
search_help_articles
Emails an existing invoice to its customer, optionally with a note and extra recipients on cc. This reaches a real inbox rather than previewing anything. A draft invoice sends as-is; to have it go out as an issued invoice, publish it first with update_invoice (publish: true) before sending.
send_invoice
The Finaloop app's "Change accounting treatment" action on the inventory purchases page: moves inventory purchases between Inventory (shown as Inventory A/P under unit tracking), COGS and Dropshipping A/P, for periods on a sales-based inventory method. The inventoryPurchases query lists the purchases with each category line's serialNumber and current type; a purchase changes as a whole or line by line. Each purchase comes back CHANGED, UNCHANGED (already that treatment) or REJECTED with a reason — a closed books period, a manually entered or system-generated purchase, a bill credit (its treatment follows the credit's reason), a category that tax rules keep out of COGS, a default rule that keeps the line in COGS, a dropshipping limit, or the purchase changing while the request ran — and the rest of the call still goes through. An unexpected error stops the call: that purchase comes back FAILED (its current treatment is worth re-reading) and the purchases after it NOT_PROCESSED (untouched, safe to retry). Some changes wait for the person's go-ahead, as they do in the app: such a purchase comes back REJECTED with NEEDS_CONFIRMATION, every confirmation it still needs listed in pendingConfirmations and explained in the message — TAX_WARNING (the app's tax-advice warning for moving shipping, production or other indirect-cost lines to COGS) and PAYMENT_LINKS (moving a purchase linked to a vendor payment group to COGS may cancel that whole link group). Once the person agrees, the purchase is sent again with those values in its own confirmations; a confirmation covers only the purchase it is sent with.
set_purchase_accounting_treatment
Removes bills from a bank transaction's existing match. The payment still accounts for itself in full, so whatever the removed bills were covering is spread over the bills that remain, filling each in order up to its balance due — pass bills to set those amounts yourself instead. Removing every bill leaves the transaction with nothing to pay, so that case takes a classificationTagId for the category it falls back to.
unmatch_bill_payment
Patches an existing bill: only the fields provided change, everything else keeps its current value. Providing categories replaces the full set of categorized amounts and recomputes the bill total. Credits and adjustments stay untouched — this tool has no way to reach them, and bank-transaction links belong to match_bill_payment. To only fix the categories on a bill, categorize_bill is a shorter path to the same patch. As with create_bill, the parameter names here belong to this tool rather than to the person asking — edits arrive as "recategorize that bill as software subscriptions" or "push the due date out a week".
update_bill
Patches an existing invoice: only the fields provided change, everything else keeps its current value. Providing lines replaces the full set of line items on the invoice. Payments, voiding and deleting the invoice stay untouched — this tool has no way to reach them. To only fix the category on an invoice, categorize_invoice is a shorter path to the same patch. Passing publish: true issues a draft, requiring a due date and a category on every line in the merged state. Line categories take a few moments to settle after a categorized create or patch, so a follow-up patch made right away can still see the old categories and resend the invoice without them. As with create_invoice, the parameter names here belong to this tool rather than to the person asking — edits arrive as "push the due date out a week" or "add a shipping line".
update_invoice
Patches an existing purchase order: only the fields provided change, everything else keeps its current value. As with create_purchase_order, companies not on unit tracking (Inventory IQ) get an UNSUPPORTED_MANAGEMENT_METHOD error linking to the Inventory IQ page. Lines merge by SKU via upsertLines (new SKUs are resolved/created, existing SKUs are updated in place); removing a line is explicit via removeSkus. A new SKU that doesn't exactly match one of the company's products (after trimming surrounding spaces; case and dashes matter) creates a new whole-unit product; that product stays even if the purchase order is later deleted or the save fails, and products can't be deleted. Looking the SKU up among the company's existing products first and confirming with the person before sending a SKU that matches none avoids that; a near-identical existing SKU usually means a typo. customer and dropshipping replace in full when provided; indirectCosts replaces the order-level costs and keeps the costs tied to a shipment group. A patch keeps the order's cost allocations: one that leaves the lines and indirect costs as they are leaves the allocations untouched, an order whose indirect costs are allocated automatically is re-allocated when they change or when a draft is published, and on an order with manually allocated indirect costs publishing a draft or changing its lines or indirect costs is refused with MANUAL_ALLOCATIONS_WOULD_RESET, since it would reset those allocations — the merchant makes that change in the Finaloop app. A patch that would delete a vendor payment group linked to the order (a vendor change, a cost cut below what the group paid, or clearing the production date) is refused with PAYMENT_GROUPS_WOULD_BE_DROPPED naming the groups; this tool can't acknowledge them, so once the person agrees, the same change goes through upsertSupplyChainPurchaseOrder with those group ids in acknowledgedLinkedEntityDeletionIds, and the groups are deleted with it. Passing publish: true promotes a draft to in-progress and requires the merged state to have placementDate, inProgressDate and a directCostPerUnit on every line. Save rules checked when the order is written come back as INVALID_INPUT with the rule's code leading the message (for example INVALID_QUANTITY: …), followed by what to change. Edits that touch a closed-books (locked) period come back that way with LOCKED_PERIOD: a locked production or placement date cannot move, and a purchase order recognised in a locked period keeps its vendor, name, customer, lines, costs and draft status — notes stay editable. As with create_purchase_order, the parameter names here belong to this tool rather than to the person asking — edits arrive as "change the quantity on that SKU to 5" or "drop the shipping cost".
update_purchase_order
Finaloop ChatGPT Plugin FAQ
How the directory, categories and Discoverability Score work.
Read the methodologyHow do I improve Finaloop's ChatGPT Plugin discoverability?
The levers are the listing surface agents actually read: names, descriptions, keywords, tool metadata, and registry health. Which lever matters depends on where discovery breaks, which is what continuous measurement shows.
What are Finaloop alternatives on ChatGPT?
As of 2026-11-22, Finaloop competes with Accountable, Agenix Invoice, AgentCollect, Aikount, Akaunting, ALL-FOOD Import Calculator, AltiBooks, Ambrook and 107 more in ChatGPT Accounting & Bookkeeping, ranked by public Discoverability Score.
Where is this profile measured?
This profile uses the geography attached to the latest public registry snapshot: US. Locale tags are intentionally omitted.