MCP App Store
by KeenEthics
HelpDol icon

HelpDol MCP Server for ChatGPT

By DANIEL DE OLIVEIRA LOPES

Overview

Control your business by chatting with ChatGPT. HelpDol is an AI-powered financial management app for sales professionals, service providers, self-employed workers, freelancers, contractors, and small business owners. Connect your bank account once in the HelpDol app and your income and expenses are imported automatically — no manual typing, no duplicates. Then just ask. Check your profit, balance, cash flow, accounts payable and receivable, and how much you have paid each helper. Sort out the transfers that still need review, schedule upcoming payments with reminders, generate receipts for helpers and clients, pull your yearly report for the accountant, and charge customers by card or payment link — all from the chat.

How to connect

  1. ChatGPT: Open the HelpDol listing in the ChatGPT app directory and choose Connect. ChatGPT listing HelpDol exposes 57 tools here.

Tools

answer_ai_question

ChatGPT
Answer a question HelpDol's AI asked about the user's money (listed in pending_questions of get_ai_classifications). Today: the HOUSING question — HelpDol finds the largest payment that repeats every month (usually rent or the house payment) and asks what it is. Use when the user answers it, e.g. "sim, é meu aluguel", "é a prestação da casa", "não, isso não é moradia". The answer becomes a rule: the payment is categorized as Moradia (personal, partially deductible with a home office) now and every month. Optionally send the size of the work space to estimate the home office deduction.

audit_forecast

ChatGPT
AUDIT the HelpDol forecast of ACCOUNTS RECEIVABLE and ACCOUNTS PAYABLE: shows exactly where each expected amount comes from so the user can check it. Use it when the user says "auditar minhas contas a receber", "auditar minhas contas a pagar", "audit my forecast", "auditar mis cuentas", "this forecast looks wrong", "confere minha previsão", or doubts the expected income/expenses. Returns, for income and for expenses: the expected total until month end and its parts, every recurring item with status (due, paid/received this month, late), income sources that pay many times a month with their daily / weekly / monthly averages (Uber, Lyft, delivery apps), bills that stopped appearing (to confirm with the user), bills the user paused, the main fixed bills found or missing (housing, insurance, car payment, phone), months left out for incomplete data, and what was not counted (transfers between the user's own accounts, bank returns, refunds). All amounts are computed by HelpDol from the bank history — never recalculate them. It changes nothing; to save a correction the user confirms, call save_forecast_correction.

cancel_invoice

ChatGPT
Void an invoice that was issued by mistake (it leaves Accounts Receivable and the PDF shows VOID). Only for invoices with no bank payment linked. Always confirm the invoice number with the user first. This is NOT a way to mark an invoice as paid.

cancel_scheduled

ChatGPT
Cancel/remove a scheduled payment or income (it stops appearing and stops reminding). Use when the user says "cancela o agendamento", "não preciso mais agendar isso", "remove o lembrete". First find it with get_accounts_payable / get_accounts_receivable and confirm which one (by id).

classify_transaction

ChatGPT
Classify a bank transaction after asking the user what it was. For an INCOMING credit (type="income") the options are "recebimento_invoice" (Invoice Payment: it pays one of the user's open invoices — pass invoice_number; this is what marks the invoice PAID), "receita" (other income) or "transferencia_entre_contas". For an OUTGOING transfer the options are: classification="colaborador" (payment to a helper/employee — ALWAYS ask WHICH one and pass helper_name), "despesa" (other business expense), or "transferencia_entre_contas" (the user just moved their OWN money between their OWN accounts — this does NOT count as an expense; it is removed from income/expense totals and saved as a TRANSFER record (see get_transfers), and will not be re-imported). If it is really none of those, use "ignorar". Only use on items returned by get_pending_transactions. Always confirm with the user which transaction (show amount/date/description) before classifying.

compare_periods

ChatGPT
Compare two periods (income, expenses and profit) and show the difference and % change. Default (no args) = CURRENT month vs PREVIOUS month. Use for "como tô em relação ao mês passado", "melhorei?", "comparar este mês com o anterior".

correct_classification

ChatGPT
Correct how a bank transaction was classified (or confirm it) when the user says it is wrong — e.g. "HelpDol, change the ONELINK.TO expense to Software e assinaturas", "isso foi pessoal", "esse Zelle foi pro meu ajudante", "isso foi transferência pra minha poupança", "isso é material, não alimentação". Use EITHER id + record_table (from get_ai_classifications) OR search = the merchant/description the user mentioned (e.g. "ONELINK.TO", "QT 1023", "Stripe"); with search, every transaction from that merchant in the last 12 months is corrected (send only_this_one + amount to fix just one). The correction is saved and LOCKED, and HelpDol LEARNS it: future transactions from the same merchant are classified the same way automatically, and older ones the AI classified are updated. If the result says more than one merchant matches, show the options and ask. Categories for expenses: Combustível, Materiais, Ferramentas e equipamentos, Manutenção de veículo, Transporte, Alimentação, Supermercado, Moradia, Aluguel do negócio, Telefone e internet, Software e assinaturas, Seguros, Taxas bancárias, Impostos e licenças, Marketing, Saúde, Educação, Lazer, Vestuário, Outros. Income: Serviços ou vendas, Reembolso, Juros e rendimentos, Outros recebimentos.

create_card_payment

ChatGPT
Start a CARD payment (Tap to Pay) so the user can charge their own customer by having them tap a card on the user's phone. Use when the user says things like "quero receber 300 dólares da Joaquina", "cobrar 150 do cliente", "recebe 80 reais da Maria no cartão", "passa 200 no cartão". ALWAYS ask for the amount if it was not given, and ask who the customer is if a name would help the user recognise the payment later. This does NOT charge anyone by itself: it prepares the charge and returns a link that opens the HelpDol app on the Tap to Pay screen with the amount already filled in. ALWAYS call this tool when the user asks to receive/charge money, on ANY device — including desktop and web. You are NOT the one who opens the app: you only prepare the charge and hand over the link, which the user taps on their iPhone. NEVER reply that the action is unavailable, not supported here, or not enabled — that is wrong; the tool is available and preparing the charge is useful regardless of where this conversation is happening. After the customer taps their card, the income is recorded automatically. Tell the user to open HelpDol to finish.

create_expense

ChatGPT
Register money that ALREADY went out (e.g. "paguei 80 de gasolina"). NEVER for the future — use schedule_payment. Same rule as create_income: with a connected bank the tool asks first whether the payment left the CONNECTED account, then you call it again with through_connected_bank set. Payments made through the connected account are registered as AWAITING bank confirmation and settle by themselves.

create_income

ChatGPT
Register money that ALREADY came in (e.g. "recebi 200 do João", "entrou 50 hoje"). NEVER for the future — use schedule_payment for that. IMPORTANT: if the user has a connected bank account, this tool first answers with needs_answer=true and a question; ask the user whether the money came through the CONNECTED bank account, then call this tool again with through_connected_bank set. Money that went through the connected account is registered as AWAITING bank confirmation and settles by itself when the bank transaction arrives — this is how HelpDol avoids counting the same money twice.

create_invoice

ChatGPT
Create (issue) an invoice and return its PDF link. Turn the user's sentence into structured items, e.g. "8 hours of painting at $50/h plus $180 of material" -> items [{description:"Painting service", quantity:8, unit_price:50}, {description:"Material", quantity:1, unit_price:180, taxable:true}]. Ask ONLY for what is missing (client, what was done, amount). SALES TAX: never ask the user for a percentage and never ask "do you want to add tax?". For every invoice, find out yourself whether sales tax applies: search a reliable, current official source (state Department of Revenue / Comptroller, or the city/county rate table) for the place of the sale (the client's address; if missing, the business address) and the type of product or service (e.g. house painting in Concord, NC). If it applies, put the rate already filled in the summary before creating, e.g. "Sales tax: 7.25% — North Carolina 4.75% + Cabarrus County 2.5% (NC Department of Revenue). Keep it, change it, or no tax?" and create with that rate as soon as the user accepts or gives another one. If it does not apply (e.g. labor/services exempt in that state), say it in one line ("No sales tax: painting labor is not taxable in <state> (<source>)") and create with tax_rate 0 — do not ask. Mark exempt lines with taxable:false (labor) and taxable lines true (materials). Pass tax_rate and a short tax_description with jurisdiction and source. If the city is ambiguous (there are several Concords), ask only for the state. Before calling, show a one-line summary (client, items, subtotal, tax, total, due date) when sales tax applies (so the user can keep or change the rate) or something was ambiguous; otherwise create directly. Requires the business name (get_invoice_setup). The invoice goes into Accounts Receivable automatically; it becomes PAID only when a bank credit is linked to it — never offer to mark it paid.

delete_record

ChatGPT
Delete a record the user registered BY MISTAKE (a wrong or DUPLICATE income, expense or helper payment). Use when the user says "apaga", "exclui", "esse está errado/duplicado", "tira esse lançamento". IMPORTANT: first find the exact record with list_recent (or get_helpers) and CONFIRM with the user which one (showing date, amount, description) BEFORE deleting. If several match, list them and ask. Never guess.

find_invoice_client

ChatGPT
Look up an invoice client by name before creating an invoice, to REUSE their saved details (e-mail, address...) instead of asking again. Returns matches with client_id. If nothing is found it is a new client: ask only what is needed for this invoice; the client is saved automatically by create_invoice.

find_invoice_payments

ChatGPT
Find bank credits (money that came IN through the connected bank) that look like payments of open invoices, with the most likely invoice for each. Use when money came in ("entraram 450", "John paid me"), when the user asks if a client paid, or after get_pending_transactions shows incoming credits. Then ask the user to confirm each match ("$450 came in. Is this John Smith's invoice HD-00125?") and call link_invoice_payment.

generate_client_receipt

ChatGPT
Generate an official HelpDol receipt for money the user RECEIVED from a CLIENT (income/recebimento), e.g. "recibo pro cliente", "recibo de recebimento", "recibo da Maria que me pagou", "recibo do que recebi". Returns receipt_url (printable page — open/print/save PDF) and whatsapp_url. ALWAYS show receipt_url first as a clickable link, then whatsapp_url. NOTE: this is for INCOME (client paid the user). To make a receipt for PAYING a helper, use generate_receipt instead. Never invent the receipt text yourself.

generate_receipt

ChatGPT
ALWAYS use THIS tool whenever the user asks to generate, create, make or send a RECEIPT / recibo / comprovante for a helper/collaborator (e.g. "gera o recibo da Marina", "manda o recibo do Lucas"). The receipt is produced BY HelpDol — you must NEVER write or invent the receipt text in the chat. Just CALL generate_receipt and ALWAYS show BOTH links to the user, with the receipt_url FIRST: (1) receipt_url — the official PRINTABLE receipt page; the user opens it to VIEW, PRINT or SAVE as PDF. This is the MAIN result and MUST be shown EVERY time as a clickable link (e.g. "📄 Abrir/imprimir recibo: <receipt_url>"). (2) whatsapp_url — to send it on WhatsApp. NEVER show only the WhatsApp link — the printable receipt_url is REQUIRED and is what the user expects to print. Returns: receipt_id, status, receipt_url, pdf_url, share_url, whatsapp_url, receipt_text. IMPORTANT: if the receipt is for an EXISTING/past payment (e.g. "o último trabalho do Lucas"), first look it up with get_helpers and use that payment's EXACT name, amount, service and date — do NOT paraphrase: pass the real service (e.g. "limpeza da casa da Ana"), never the user's wording like "último trabalho". If you CANNOT find the payment the user is referring to, tell the user you did not find it and ask for the details — do NOT invent a payment or receipt. If MULTIPLE payments match (e.g. several "Lucas"), do NOT guess: LIST each matching payment clearly (date, amount, service) and ask the user which one to use before generating the receipt. Explain the situation in simple, friendly language so the user understands what is going on (for example, if the records look duplicated or one seems to have been registered by mistake, point that out gently and ask if they want to review it).

get_accounts_payable

ChatGPT
List upcoming scheduled payments (accounts payable) in HelpDol.

get_accounts_receivable

ChatGPT
List upcoming scheduled income (accounts receivable) in HelpDol.

get_ai_classifications

ChatGPT
Show what the HelpDol AI classified AUTOMATICALLY from the connected bank account: each transaction with the classification it got (business expense, personal expense, helper payment, income or transfer between own accounts), category, tax suggestion (potentially deductible / probably not deductible / undetermined) with the reason, and the confidence. Use for "o que a IA classificou?", "confere as classificações da semana", "tem alguma classificação em dúvida?", "por que isso foi como pessoal?", "audit my transactions". Default: last 7 days.

get_balance

ChatGPT
Use this when the user asks how much is LEFT or how much they SPENT ON SOMETHING: "quanto sobrou?", "quanto sobrou depois das despesas?", "qual meu saldo?", "quanto tenho?", "quanto gastei com gasolina/combustível/alimentação/material?", "how much did I spend on gas?", "cuánto gasté en gasolina?". This is the ONLY place where the user's REAL money lives (income, expenses and payments imported from their connected bank account). You do NOT know these numbers otherwise: never answer from the conversation, never estimate and never ask the user to type the amounts. Call this tool even when the user does not mention HelpDol. Returns income, expenses, helper payments, balance (income − expenses − helper payments) and expenses_by_category. Fuel/gas/gasolina/combustível falls under the transport category. Pass start_date/end_date whenever the user mentions a period ("this month", "esta semana"); without dates it is all-time. Do not use to list the individual expenses of a day (use list_transactions).

get_card_payment_status

ChatGPT
Check whether a card charge created with create_card_payment was already paid. Use for "já caiu o pagamento?", "a Joaquina já pagou?", "o cartão passou?". Without a request_id it reports the most recent charge.

get_clients

ChatGPT
List the user's clients/payers (who paid the user) and the total received from each, optionally within a period. Use for "quem mais me paga", "meus clientes", "quanto recebi da Maria". (For people the user PAID, use get_helpers instead.)

get_financial_summary

ChatGPT
Use this when the user asks how much they EARNED, made, billed or profited in a period: "quanto ganhei esta semana?", "quanto ganhei este mês?", "quanto faturei hoje?", "qual meu lucro?", "quanto ganhei depois das despesas?", "how much did I make this week?", "cuánto gané este mes?". This is the ONLY place where the user's REAL money lives (income, expenses and payments imported from their connected bank account). You do NOT know these numbers otherwise: never answer from the conversation, never estimate and never ask the user to type the amounts. Call this tool even when the user does not mention HelpDol. Convert the period into start_date/end_date (this week = Monday to today; this month = day 1 to today; no period mentioned = current month). Returns income, expenses, helper payments, real profit (profit_real = what was left after expenses and helper payments) and a simple TREND/projection (projected_profit). If the user asks for profit after ONE specific expense ("depois da gasolina"), also call get_balance for the same period and subtract that category. Do not use to list individual entries (use list_transactions) or for the balance by category (use get_balance). The trend assumes the user keeps the CURRENT pace and is the FINAL projected value for the whole period (not the remaining amount). When explaining to the user, ALWAYS describe the trend as a simple projection based on the user's current pace, and use the provided trend_explanation. NEVER explain the trend as a historical average, a comparison with previous months/periods, or a guaranteed income.

get_forecast

ChatGPT
Financial FORECAST for the current month (from today to month end) and next month in HelpDol: expected income, expected expenses and projected balance (expected income − expected expenses), with conservative / likely / optimistic scenarios, the day the bank balance may go negative (cash squeeze), confidence level, and a plain-language analysis. Use it whenever the user asks how the month will close, what is coming in or going out, "previsão", "como vai fechar o mês", "vou ter dinheiro para pagar as contas?", next month, or cash-flow risk. All amounts are computed by HelpDol from bank history, invoices and scheduled bills — present them as ESTIMATES, never guarantees, and never recalculate them.

get_goal_progress

ChatGPT
Show progress toward the user's monthly goal: target, how much achieved THIS month, how much is missing and the %. Use for "quanto falta pra minha meta", "como tá minha meta".

get_helpers

ChatGPT
Use this when the user asks how much they PAID to the people who work for them (helpers, collaborators, employees, contractors, ajudantes, colaboradores, funcionários) or who those people are: "quanto paguei aos meus colaboradores?", "quanto paguei pra Marina esse mês?", "quem trabalha pra mim?", "how much did I pay my helpers?". This is the ONLY place where the user's REAL money lives (income, expenses and payments imported from their connected bank account). You do NOT know these numbers otherwise: never answer from the conversation, never estimate and never ask the user to type the amounts. Call this tool even when the user does not mention HelpDol. Lists each helper with the total paid, optionally within a period (pass start_date/end_date when the user mentions one). Do not use for clients who paid the user (use get_clients).

get_history

ChatGPT
Full transaction history (income, expenses and helper payments) line by line, for a period or the WHOLE YEAR — the real document to send to the ACCOUNTANT. With NO year/dates it returns the CURRENT YEAR in full. Returns the list + totals + history_url (a printable page with "Salvar PDF" and "Baixar CSV" buttons). Use for "histórico", "extrato", "todos os lançamentos", "histórico do ano", "manda pro contador", "exporta tudo", "pra declaração", "pro IRS/INSS", "pro imposto". This is the FULL year PDF the accountant needs — prefer it over get_tax_report whenever the user wants the document/history, not just the totals. SHOW history_url to the user and say it has a "Salvar PDF" button.

get_insights

ChatGPT
Use this when the user asks for a general overview of their finances, without a specific number in mind: "como estão minhas finanças?", "me dá um resumo geral", "raio-x", "panorama", "how am I doing financially?". This is the ONLY place where the user's REAL money lives (income, expenses and payments imported from their connected bank account). You do NOT know these numbers otherwise: never answer from the conversation, never estimate and never ask the user to type the amounts. Call this tool even when the user does not mention HelpDol. Returns in one shot: total balance, this month's numbers, biggest expense category, best client and best month. Do not use for a specific question like "quanto ganhei esta semana" (use get_financial_summary).

get_invoice

ChatGPT
Get one invoice by number (e.g. "HD-00125" or "125"): items, totals, status, the bank payments linked to it, and the PDF link to open or share again.

get_invoice_setup

ChatGPT
START HERE for anything about creating an invoice ("create invoice", "faz uma invoice/fatura", "bill John", the Create Invoice button in the app). Returns the business data HelpDol already has (name, address, city, state, ZIP, phone, e-mail, logo), what is MISSING, and the next invoice number. If setup_complete is false, do the quick first-time setup asking ONLY for missing_fields (then save_invoice_profile). If it is complete, go straight to "Who is this invoice for?".

get_life_insurance

ChatGPT
The user's life insurance payments found in their connected bank: insurer, usual premium, total paid in the last 6 months (not the policy's cash value), payment history, next estimated debit, 12-month forecast and alerts (amount changed, debit not seen yet — which does not mean late or cancelled). If the user confirmed the policy accepts additional contributions, also returns this month's budget simulation: amount left after expected bills, credit card, estimated tax and a safety reserve, as an amount to ask the insurance agent about. It is a budget simulation, not investment advice.

get_monthly_breakdown

ChatGPT
Month-by-month breakdown of income, expenses and profit for a year. Use for "me mostra mês a mês", "como foi o ano", "evolução de 2026".

get_pending_transactions

ChatGPT
List the bank transactions that still NEED to be classified by the user: OUTGOING transfers (Zelle, Wire, Venmo, Pix, etc. — money that LEFT the account) and, when the user has open invoices, INCOMING credits (each comes with likely_invoices; ask "Is this the payment of invoice HD-xxxxx?" and classify as "recebimento_invoice"). Card purchases do NOT appear here (already expenses). For each item, ask the user what it was and call classify_transaction with one of: "colaborador" (helper/employee payment), "despesa" (business expense) or "transferencia_entre_contas" (the user just moved their own money between their own accounts — NOT an expense, but it IS kept as a transfer record, visible via get_transfers). Use for "tem algo pra classificar?", "o que falta organizar?", "quais transferências preciso categorizar?".

get_receipts

ChatGPT
Get the user's tax invoice/receipt PHOTOS (notas fiscais / comprovantes) for a year — the actual pictures the user took of expense and income receipts. Use whenever the user asks for "notas fiscais", "comprovantes", "as fotos das notas", "manda as notas pro contador", "documentos do IRS". Returns: count, total_value, gallery_url (a printable gallery with ALL photos to save as PDF and send to the accountant) and receipts[] with each photo url, date, amount and description. SHOW gallery_url to the user. NEVER invent receipts — only show what is returned. If count is 0, gently tell the user they can photograph receipts in the app.

get_recurring_payments

ChatGPT
Upcoming RECURRING bills/charges that HelpDol detected from the user's CONNECTED BANK HISTORY (subscriptions, fixed bills, utilities, memberships, etc.). Use this whenever the user asks what they have to PAY: "quais contas eu tenho para pagar esta semana?", "quanto vou pagar até o dia 18?", "o que tenho para pagar este mês?", "quanto gasto por mês com contas recorrentes?". It analyzes the repeating charges in the bank transactions, predicts the NEXT occurrence of each one (next_due) and returns them filtered by the window you ask for, plus monthly_recurring_total (how much the user spends per month on recurring bills). All amounts are estimates based on the bank pattern — answer only from the returned list; do not invent bills. This reads ONLY recurrences detected/confirmed from the bank history.

get_tax_report

ChatGPT
Generate an annual tax/IRS/INSS report for a given year: total income, expenses, helper payments, deductible expenses and net profit (a quick SUMMARY), plus report_url (printable) AND full_history_url (the COMPLETE year history, line by line, with a "Salvar PDF" button — the actual document for the accountant). Use for "relatório do IRS/INSS", "relatório anual", "quanto vou declarar". ALWAYS show BOTH links and offer the full history (full_history_url), not just the summary. If the user clearly wants the full history/extract itself, prefer get_history.

get_transaction

ChatGPT
Get the full details of ONE specific transaction (income, expense or helper payment) by its id. Use after list_transactions / get_pending_transactions when the user asks about a specific entry ("essa transação do Zelle", "detalhe desse lançamento"). Returns amount, date, description, category, source (bank/manual), whether it still needs classification and whether it has a photo attached.

get_transfers

ChatGPT
List the money the user MOVED BETWEEN THEIR OWN ACCOUNTS (transfers classified as "transferencia_entre_contas"). These are NOT income and NOT expenses — they are just the user moving their own money, kept as a separate record. Use for "quanto transferi entre minhas contas?", "quais foram minhas transferências?", "me mostra as transferências entre contas", "o que eu movi de uma conta pra outra". Do NOT use this for income/expense questions — use get_financial_summary or list_transactions for those. Defaults to the CURRENT YEAR when no period is given. NEVER answer this question from memory or by reasoning over a previous list of pending transactions — a transfer only counts once the user has actually classified it. If this tool returns zero, say plainly that nothing has been classified as a transfer between the user's own accounts in that period, and offer to look at what is still pending with get_pending_transactions.

irs_calculate

ChatGPT
"Calculate my income tax": runs the HelpDol tax engine (official IRS rules of the tax year) on HelpDol data + interview answers + expense review, saves the result and returns: tax year, taxable business income, deductible expenses, self-employment net profit, deductions, taxable income, federal tax, self-employment tax, state tax, credits, withholding and estimated payments, the estimated amount to pay or refund, how much the deductions really reduced the tax, joint vs separate scenarios and the PDF link for the accountant. It is an ESTIMATE, not a filed return. Never recompute the numbers.

irs_get_expense_review

ChatGPT
The mandatory "Review expenses for the IRS" step: bank expenses of the tax year (grouped by merchant, e.g. "Verizon — 12 payments, $1,140") that the AI is NOT sure are business: internet, phone, fuel, Home Depot, meals… Each comes with options Business / Personal / Mixed / Not sure. Undecided ones stay OUT of the tax calculation. Saved decisions are reused next year.

irs_get_report

ChatGPT
"Create a report for my accountant": PDF with the tax information provided, income summary, deductible expenses by category, home office and vehicle calculations, credits and deductions considered, tax estimate, expenses pending confirmation and the methodology / tax-year rules. Preparation for the accountant, not a return filed with the IRS.

irs_get_simulation

ChatGPT
Get the user's U.S. income tax (IRS) simulation for a tax year: status (not started, in progress or calculated), what HelpDol already knows from the user's records (income, payers, rent and utilities, tax payments), the answers saved so far, the next questions to ask and the previous result. Use when the user wants to simulate their income tax, review deductible expenses, calculate a home office deduction, see a previous simulation or prepare a report for their accountant. Covers individuals with W-2 wages and/or self-employment (Schedule C). It is an estimate, not a filed return.

irs_review_expenses

ChatGPT
Save the user's decisions from "Review expenses for the IRS". decision: business | personal | mixed | unsure. For mixed, business_percent (1–99) and a short justification the user CONFIRMED are required (e.g. phone 60% business: "used for client calls and scheduling"). It never changes the bank history and never makes an expense deductible without the user deciding.

irs_save_answers

ChatGPT
Save the answers the user gave in the income tax (IRS) simulation interview, so they can stop and continue later. Send only the sections that changed; they are merged with what was saved before (null clears a field; dependents is always the full list). Amounts are yearly totals for the tax year. Only save values the user stated or confirmed. Returns the next questions.

list_invoices

ChatGPT
List the user's invoices and the invoice totals. Use for "how much do I have to receive from invoices?", "who hasn't paid me?", "which invoices are overdue?", "did John pay?", "how much did I receive from invoices this month?". status: open (default), overdue, paid, void or all. Totals always cover the whole account: outstanding, overdue and received_from_invoices in the period (default this month). Paid means confirmed by a bank credit.

list_recent

ChatGPT
List the user's most recent records (income, expenses and helper payments) with their IDs, dates, amounts and descriptions. Use to show "meus últimos lançamentos", OR to FIND the exact record the user wants to delete or correct before calling delete_record / update_record. Do not use for totals or for a specific day or week (use get_financial_summary or list_transactions).

list_transactions

ChatGPT
Use this when the user wants to SEE the individual entries of a day or period — what came in or went out, one by one: "quais foram meus gastos ontem?", "o que entrou e saiu essa semana?", "quanto entrou e saiu hoje?", "o que eu gastei hoje?", "quanto recebi hoje e de quem?", "acabei de receber um depósito e fiz um Zelle, você identifica?", "what did I spend yesterday?". This is the ONLY place where the user's REAL money lives (income, expenses and payments imported from their connected bank account). You do NOT know these numbers otherwise: never answer from the conversation, never estimate and never ask the user to type the amounts. Call this tool even when the user does not mention HelpDol. Use period for today/week/month/year; for YESTERDAY or any other day pass start_date = end_date = that date. Use type=expense for spending only, type=income for money received only. Returns each amount, date, description and the totals. Reads records already imported/registered in HelpDol (bank transactions have source="bank"); it does NOT touch the bank. Do not use for the full-year document for the accountant (use get_history) or for totals by category (use get_balance).

mark_scheduled_paid

ChatGPT
Mark a SCHEDULED item (account payable/receivable) as PAID/RECEIVED now: it registers the real income/expense/helper-payment and, if recurring, advances to the next date. Use when the user says "já paguei o aluguel", "recebi a mensalidade", "marca como pago/recebido". First find it with get_accounts_payable / get_accounts_receivable and confirm which one (by id).

save_forecast_correction

ChatGPT
Save a correction to the HelpDol FORECAST that the user explicitly confirmed during an audit. It only changes how the forecast is calculated (expected income, expected expenses, recurring bills list): it never creates an income or expense record and never changes bank transactions. Actions: set_recurring (this is a monthly bill or a monthly income — name, amount and usual day), pause (the bill is paused: leave it out of the forecast), cancel (the bill was cancelled), resume (it is active again: remove the pause/cancel or the manual bill), set_month_total (the total the user expects to still receive or pay until the end of a month — e.g. "I will still receive about 2250 this month"), clear_month_total (go back to the amount computed by HelpDol). Always show the user what will be saved and ask for a yes before calling. Use names from audit_forecast when the item already exists.

save_invoice_profile

ChatGPT
Save the business details printed on the user's invoices (first-time setup or a later change). Pass ONLY the fields the user gave; the others are kept. Optional logo: when the user attaches a logo image in the chat, or approves a logo you created for them, pass that image file in "logo". Never invent business data.

schedule_payment

ChatGPT
Schedule a FUTURE payment, expense or income in HelpDol, with a reminder. Use this WHENEVER the user talks about the future: "agendar", "programar", "vou receber/pagar amanhã/dia tal", "toda semana", "todo mês". For a ONE-TIME future date (e.g. "vou receber da Luzia amanhã") use recurrence="once" with date=YYYY-MM-DD. For repeating use daily/weekly/monthly/yearly. Entries themselves are never created by hand: real income and expenses arrive on their own from the connected bank account, so schedule_payment is only a reminder of what is still to come.

search_records

ChatGPT
Search the user's records (income, expenses, helper payments) by a word/name — useful to find an OLD record (not just the recent ones) to view, correct or delete. Use for "acha o pagamento da Marina", "procura lançamentos de gasolina", "cadê o recibo do Lucas". Returns matches with their ids. Do not use to answer how much was spent on a category (use get_balance).

set_goal

ChatGPT
Set the user's MONTHLY goal: an income target or a profit target. Use for "minha meta é receber 5000 por mês", "quero lucrar 3000 esse mês", "definir meta".

update_life_insurance_policy

ChatGPT
Save what the user confirmed about a life insurance policy: policy type, whether it accepts additional contributions (only with the source the user confirmed: their agent or the policy document), the agent's name, phone and email, or that the payments are not life insurance. Only save what the user stated.

update_record

ChatGPT
Correct/fix an existing record: change the amount, description, name (client/helper), service or date. Use when the user says "corrige", "muda o valor para", "o nome certo é", "a data era". First find the record with list_recent and confirm which one, then update only the fields the user wants to change.

Capabilities

WritesInteractive

Example Prompts

Click any prompt to copy it.

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/helpdol"><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/helpdol)