accounting_report
ChatGPTOne of the standard finance reports for this organisation. "trial_balance" and "balance_sheet" are computed LIVE from every journal line — a voided entry and its reversal cancel out — so they reflect a posting immediately. "profit_and_loss" and "cash_flow" are per calendar month and need a year and a month. "ar_ageing" and "ap_ageing" bucket what is outstanding by how overdue it is. "budget_vs_actual" needs a year. "overview" is the dashboard summary for one month.
acknowledge_review
ChatGPTConfirm that you have read and received YOUR OWN submitted performance review. Only the employee being reviewed can do this — it cannot be done on somebody else's behalf. Answers a preview of what you would be acknowledging plus a confirm_token first.
add_shift_note
ChatGPTAttaches a note to one shift assignment, recorded against the connected admin as its author. Notes are what the next person on that shift reads at handoff. 1-2000 characters.
adjust_attendance_punch
ChatGPTRecord a punch an employee missed — a forgotten clock-in, a break that was never punched. The time is wall clock in the employee's own attendance timezone. The punch is bound to the matching scheduled shift and the day is re-finalized, so worked hours and payroll follow automatically. Answers a preview first; nothing is written until you send the confirm_token back. For SEVERAL punches for one employee on one day — a break is two (break_start and break_end) — use adjust_attendance_punches, which needs only ONE confirmation.
adjust_attendance_punches
ChatGPTRecord up to 6 missed punches for ONE employee on ONE day under a single confirmation — use this for a break (break_start and break_end together), or for a forgotten clock-in and clock-out. Times are wall clock in the employee's own attendance timezone. All punches are recorded together or none is, and the day is re-finalized once. Answers a preview listing every punch and the resulting break and worked time; nothing is written until you send the confirm_token back.
approve_expense
ChatGPTApprove one submitted expense. Answers a preview — amount, vendor, description, which approval level this is and whether it is the last one — plus a confirm_token. On the final level Staffology posts the journal entry, which is accounting state. You cannot approve an expense you created yourself.
approve_leave_request
ChatGPTApprove one pending leave request. Answers a preview first — who, which leave type, the dates, the days deducted and their remaining balance — plus a confirm_token; only a second call echoing that token approves it. Approving draws the days off the balance and marks them as leave in attendance, and cannot be undone from here.
assign_employee_location
ChatGPTPoints one or more people at a site. Staffology holds this in two places — the person's home office and the site their attendance is geofenced against — and by default this sets BOTH, which is what keeps them from drifting apart. Leaving the location out unassigns the people named. Every name has to resolve to somebody in this organisation or the whole call is refused; there is no partial assignment.
assign_shift
ChatGPTPuts a single shift on one person's calendar on one date — the per-cell equivalent of the scheduler grid, and the only way to set custom times or a note on a shift. Give a template to inherit its times, or give the times directly, or both (the times you give win). The shift is created as a DRAFT: nobody is notified and the employee sees nothing until publish_shifts runs. A second shift on the same day at a DIFFERENT start time is a legal split shift and is allowed; the same start time is refused.
assign_shift_swap_target
ChatGPTSets the employee who would take the shift on a swap request that is still pending. Nothing moves yet — approve_shift_swap is what performs the swap. The target must be an active employee in this organisation and cannot be the person who asked.
attendance_compliance_export
ChatGPTPer-employee scheduled hours, worked hours, overtime, late days, absent days and days on leave over a period, in the shape a labour-department audit asks for, returned in this answer (no file or link is created). THIS SPENDS ONE OF THE ORGANISATION'S MONTHLY REPORT EXPORTS, shared with every other export in Staffology, so do not call it speculatively. The period may span at most 366 days.
attendance_trends
ChatGPTWeek-over-week or month-over-month comparison of attendance, hours and task completions.
bulk_assign_shifts
ChatGPTCreates draft shift assignments from one template, for the given employees, on the chosen weekdays within a date range. Never overwrites an existing shift: a cell that is already taken is reported as a conflict and left alone — to REPLACE existing shifts, use overwrite_shifts instead. The shifts are created as DRAFTS — nobody is notified until publish_shifts runs. The preview states exactly how many shifts would be created and how many cells already have one.
check_shift_compliance
ChatGPTAnswers which of the shifts you are about to schedule would break a labour rule — too many days in a row, or too little rest before or after — WITHOUT creating anything. Takes exactly the arguments bulk_assign_shifts takes, so you can run it first and show the warnings. It compares the proposal against shifts that already exist, so the answer is about the roster as it would actually be. An organisation that has set no rules gets an empty answer, which means "no rules to break", not "no problems".
create_account
ChatGPTAdds one account to the chart of accounts. The code has to be unique within the organisation. The TYPE decides the sign of the balance this account contributes to every report, and it can never be changed once the account carries postings, so choose it deliberately.
create_accounting_contact
ChatGPTAdds one customer or one vendor. Names are NOT unique, so calling this twice with the same name creates two contacts — check list_accounting_contacts first. Credit limit applies to customers only and bank details to vendors only; supplying the wrong one is refused rather than ignored.
create_bank_account
ChatGPTAdds one bank account. Point it at a general-ledger account by code so that cash postings through it land in the right place; without one, cash resolves through the organisation default instead. Marking it the default demotes any other default.
create_employee
ChatGPTCreate a new employee: a Staffology login plus their roster record. Requires the gpt.write scope, the organisation's assistant write opt-in, and confirmation — the first call answers a preview (seat usage, whether the email is already taken, the exact record that would be created) and a token to echo back. No password is returned; the new employee signs in with "Forgot password", or an admin sets one in Staffology. Pay is NOT set here — set it in Staffology after the employee exists.
create_office_location
ChatGPTAdds a site people can be based at and clock in from. The coordinates and radius are the geofence: a punch is matched to this location when it happens inside that circle, so a radius that is too small will reject real arrivals and one that is too large will accept a punch from the car park next door. It is active at once and clock-in accepts ANY active site, so every employee can clock in there immediately. Subject to the work-site limit.
create_shift_template
ChatGPTAdds a shift pattern the scheduler can assign from. Creating a template schedules nobody — it is the pattern, not the roster. An overnight shift is fine: give an end time earlier than the start time and Staffology treats it as finishing the next day.
create_task
ChatGPTCreate a task and optionally assign it. Requires the gpt.write scope, and the organisation must have ChatGPT writes enabled in Settings.
create_tax_code
ChatGPTAdds one tax code with its rate. Point it at the account collected tax is owed from and, for a purchase tax, the account recoverable input tax is claimed through — without the recoverable account, input tax on an expense is folded into the expense itself rather than split out. Making it the default demotes any other default.
daily_report_pdf
ChatGPTAttendance overview plus a task-fulfilment table as a PDF. Returns a download link valid for 2 hours that ANYONE holding it can open, so share it only with people allowed to see this data. Spends one of the organisation's monthly exports.
detailed_report_pdf
ChatGPTEverything in the summary PDF plus per-employee attendance timelines, task completions and pending work. Returns a download link valid for 2 hours that ANYONE holding it can open, so share it only with people allowed to see this data. Spends one of the organisation's monthly exports.
dismiss_attendance_anomaly
ChatGPTMark one flagged attendance anomaly as reviewed and drop it from the anomalies list for every admin in the organisation; it stays dismissed when the list is recomputed. The punches themselves are not changed. Answers a preview of the anomaly first; nothing is dismissed until you send the confirm_token back.
employee_activity
ChatGPTAttendance punch timeline, day summary, task completions and active assignments for a single employee on a given date.
employee_monthly
ChatGPTTotal hours, days present vs absent, average hours/day, tasks completed and a per-day breakdown for one employee.
find_shift_coverage
ChatGPTFor one shift, lists every other active colleague and says whether each is available, already on another shift that overlaps it, or on approved leave that day. Use it when somebody calls in sick, then assign_shift_swap_target or update_shift_assignment to act on the answer. It does not change anything.
get_accounting_settings
ChatGPTThe accounting configuration for this organisation: base currency, which month the fiscal year starts in, accrual or cash reporting, the invoice and bill number prefixes and the next number in each sequence, the default tax code and bank account, whether expenses need approval, and which account code cash posts to.
get_journal_entry
ChatGPTOne journal entry with every debit and credit line, the account each line hits and the totals on both sides. Read this before voiding an entry — it is what a human needs in order to recognise the WRONG entry.
get_scheduler_snapshot
ChatGPTShifts, templates, pending swap requests and the labour rules for a date range, in a single answer — the same call the scheduler screen makes when it loads. Use it instead of four separate reads when you need the whole picture. If the range holds more than 5000 shifts the list is cut short and the answer says so explicitly; narrow the range rather than trusting a truncated roster.
get_shift_rules
ChatGPTThe organisation's labour rules: maximum hours a week, minimum rest between shifts, maximum consecutive days, whether overtime is allowed, whether swaps need an admin, and how many days ahead shifts auto-publish. These are the rules Staffology enforces when a shift is saved and warns about when a swap is approved. Answers the built-in defaults, clearly marked, when the organisation has never set them.
grant_shift_override
ChatGPTLets somebody do something on one shift that the normal rules forbid — most often clocking in outside the allowed window. Takes effect immediately, and stays in effect until it expires or revoke_shift_override withdraws it. A reason is required and is recorded against the admin granting it. Granting the same override twice creates two, so check the shift first.
import_bangladesh_holidays
ChatGPTAdds the ten Bangladesh public holidays for 2026 to the organisation calendar as non-working days: Language Martyrs' Day, Independence Day, Pahela Boishakh, both days of Eid al-Fitr, both days of Eid al-Adha, National Mourning Day, Durga Puja and Victory Day. Holidays already on the calendar for the same date and name are left exactly as they are, so running it twice is safe and adds nothing. Only 2026 — later years are not covered.
late_arrivals
ChatGPTWho arrived late, measured against each person's scheduled shift start plus the shift's grace period; a day with no scheduled shift is not counted as late. Repeat offenders highlighted.
launch_review_cycle
ChatGPTActivate a draft review cycle and create the review forms it calls for — a self review and/or a manager review per eligible employee, according to the cycle's own settings. A BULK operation: the preview states which cycle, which employees are eligible and how many review forms would exist afterwards, and nothing runs until a second call echoes the confirm_token.
leave_balance_summary
ChatGPTHow much leave each employee has and has used. Staffology keeps TWO leave ledgers and this returns both, labelled: "entitlements" is the per-leave-type allowance that approving a request draws down, and "accrual_balances" is the accrual-policy ledger that adjust_leave_balance and run_leave_accruals move. Each accrual row carries the policy_name adjust_leave_balance takes.
list_accounting_audit_log
ChatGPTThe accounting audit trail: which table and row changed, whether it was an insert, update or delete, when, and by whom. Ten tables are covered by the trigger that writes it — accounts, bank accounts, bills, budgets, credit notes, customer payments, financial transactions, invoices, journal entries and vendor payments — so a change to anything else will not appear here.
list_accounting_contacts
ChatGPTThe customers or the vendors of this organisation with their contact details, currency, payment terms and (for customers) credit limit. This is the lookup invoices, bills, payments and credit notes select a party from.
list_accounting_documents
ChatGPTSales invoices, purchase bills or credit notes with their number, party, dates, totals, how much has been paid, how much is still outstanding and their status. Amount outstanding is a generated column, so it always agrees with the row. This is the lookup payments, voids and deletions select a document from.
list_accounting_payments
ChatGPTMoney received from customers or paid to vendors: the date, amount, method, reference, bank account, and which documents each one was applied to. Note that Staffology does not post a journal entry for a payment — the ledger side is a separate manual entry — so a payment here settles the invoice or bill without moving the general ledger.
list_accounting_periods
ChatGPTThe accounting periods on record with their status, when each was locked and by whom, and when and why any was reopened. A month with NO row is OPEN. A LOCKED month refuses every posting, edit and deletion whose entry date falls inside it. Periods auto-lock seven days after month end.
list_accounting_reference
ChatGPTOne of the small reference tables the rest of the module selects from: cost centres (which a journal line can be tagged with), payment methods, or budgets (a planned amount per account per period, which budget_vs_actual compares against).
list_accounts
ChatGPTEvery general-ledger account in this organisation: its code, name, type (asset, liability, equity, revenue, cogs or expense), sub-type, whether it is a system account the automatic postings depend on, and whether it is still active. This is the lookup every journal line and every budget selects against — an account is named by its CODE.
list_bank_accounts
ChatGPTThe organisation bank accounts: name, bank, last four digits, currency, opening balance, which general-ledger account each one posts to, whether it is the default, and whether it is still active. Cash postings resolve through this — never through a hardcoded account code.
list_bank_statement_lines
ChatGPTStatement lines imported for one bank account, with the date, description, amount, running balance and — where one has been matched — the journal entry it points at. This is the lookup match_bank_statement_line selects from, and the unmatched ones are what stop a reconciliation being completed.
list_employees
ChatGPTEvery employee in the signed-in admin's organisation, with role and admin flag.
list_expenses
ChatGPTExpense transactions for this organisation with amount, currency, date, description and approval status. Defaults to the SUBMITTED queue — the ones waiting for approval. Each row carries the transaction_id that approve_expense / reject_expense / void_expense take.
list_journal_entries
ChatGPTJournal entries in date order with their description, reference, type and where they came from. Voided entries are EXCLUDED by default, and so are their reversals: a void leaves the original (flagged void) and a balanced reversing entry in the ledger, which cancel each other, so the default list shows only entries still in force. Use get_journal_entry for one entry line by line.
list_leave_requests
ChatGPTLeave requests for this organisation with the employee name, leave type, dates and day count. Defaults to the PENDING queue — the ones waiting for a decision. Each row carries the request_id that approve_leave_request / reject_leave_request / cancel_leave_request take, so list first and never guess an id.
list_office_locations
ChatGPTThe organisation's sites — name, address, coordinates, geofence radius, timezone and whether each is in use — with the location id the other location actions need. Also counts how many people are based at each. In Staffology a "work site" and an "office location" are the same thing under two names.
list_org_holidays
ChatGPTThe organisation's holiday calendar, with the holiday id the update and delete actions need. Includes whether each is a working day and whether it recurs annually. Give a year or a date range and ANNUAL holidays are projected into it — a holiday entered once in 2024 and marked as recurring is listed on its 2026 date too, flagged as a projection, so "is this roster week a holiday" is answerable. With neither, the raw calendar entries are listed as stored.
list_pay_grades
ChatGPTThe pay grades defined for this organisation — name, code, base salary, piece rate, overtime multiplier, whether the grade is active, and how many employees are on it. This is the lookup set_pay_grade selects against.
list_performance_reviews
ChatGPTPerformance reviews in this organisation with the employee, the reviewer, the cycle, the review type, the status and the overall rating. PRIVATE REVIEWER NOTES ARE NOT RETURNED. Each row carries the review_id acknowledge_review takes.
list_reconciliations
ChatGPTReconciliations for one bank account: the statement date, the statement balance, the book balance, the variance between them (a generated column, so it always agrees), the status, and who reopened one and why.
list_salary_advances
ChatGPTSalary advance requests in this organisation with the employee, the amount, the monthly repayment deduction and how many months it runs for. Defaults to the ones still pending a decision. This is the lookup approve/reject/disburse select against.
list_shift_assignments
ChatGPTScheduled shifts in a date range, with the assignment id each write action needs. Optionally narrowed to one employee by name, or to unpublished (draft) shifts only.
list_shift_swaps
ChatGPTShift swap requests with the swap id the approve, assign-target and cancel actions need, plus who asked, whose shift it is, and what state the request is in.
list_shift_templates
ChatGPTThe shift patterns this organisation schedules from — name, start and end time, break, and the template id every scheduling action needs. Optionally also counts, per template, how many already-scheduled shifts are still on the template's OLD times, which is what reapply_template_times would fix.
list_task_corrections
ChatGPTRequests from employees to correct a task completion record, with who asked, which task, which day, what they are asking for and why. Defaults to the PENDING queue. Each row carries the request_id review_task_correction takes.
list_tasks
ChatGPTActive tasks with their assignees, optionally filtered by priority.
list_tax_codes
ChatGPTThe tax codes defined for this organisation: code, name, rate percent, whether it is a sales or purchase tax, which accounts collected and recoverable tax post to, and which one is the default. Invoice and bill lines name a tax code by its CODE.
list_tax_return_periods
ChatGPTTax return periods with the tax collected, the tax paid, the net due (a generated column), the status and the filing reference. A FILED period is frozen — its figures are what was submitted — and only marking it paid is allowed afterwards.
logout
ChatGPTImmediately revokes the caller's own assistant connections to Staffology in this organisation — every ChatGPT or MCP client they have connected, not just this one. The next request will ask them to sign in again. Use when the user says "log me out", "logout" or "disconnect".
overtime_summary
ChatGPTOvertime hours and estimated cost per employee for a month, calculated at 1.5x.
payroll_period_summary
ChatGPTThe state of one payroll month — whether it exists, its status (draft, review, approved or paid), how many entries it has, the totals for gross pay, additions, deductions and net pay, which employees are still missing hours or days, and whether the month is locked against attendance edits. Read this BEFORE any payroll write: it is where the amounts in every confirmation preview come from.
productivity_ranking
ChatGPTEmployees ranked by task completion rate, total completed and tasks per hour.
publish_shifts
ChatGPTMakes draft shifts visible to employees and notifies each person whose schedule changed. Give EITHER a date range (publishes every unpublished, non-cancelled shift in it) OR a list of assignment ids from list_shift_assignments. This cannot be undone from here. The preview states how many shifts and how many people would be notified.
red_flags
ChatGPTRecent anomalies: high absence rates, zero task completions, overdue tasks, unusually short days, no activity today. Severity-ranked.
reject_expense
ChatGPTReject one submitted expense, recording the reason. Answers a preview plus a confirm_token first. Nothing is posted to the ledger.
reject_leave_request
ChatGPTReject one pending leave request. Answers a preview plus a confirm_token first. A rejected request cannot be reopened — the employee would have to ask again — so the reason is worth giving.
review_attendance_correction
ChatGPTDecide one employee's pending attendance correction request for a given day. The employee is notified either way. Approving records the decision; it does not by itself move any punch. Answers a preview of what the employee asked for first; nothing is decided until you send the confirm_token back.
review_attendance_regularization
ChatGPTDecide one employee's pending regularization for a given day. APPROVING WRITES THE REQUESTED PUNCHES and re-opens that day for finalization, so it changes worked hours and pay. Answers a preview of the exact times that would be recorded first; nothing is written until you send the confirm_token back.
review_task_correction
ChatGPTDecide one pending task correction request. Approving writes the completion record the employee asked for; either way they are notified. Answers a preview — who, which task, which day, the current record and what it would become — plus a confirm_token first.
set_work_week
ChatGPTSets the organisation's non-working days — Friday and Saturday in Bangladesh, Saturday and Sunday in much of Europe, and often just one day. This is what attendance uses to decide whether a day was expected to be worked, so changing it changes how existing days are judged, not only future ones. At least one day of the week has to stay a working day. The preview shows the current weekend and the new one.
shift_coverage_summary
ChatGPTHow many people are scheduled on each day of the range, how many of those shifts are still drafts, and which days have NOBODY scheduled at all. The uncovered days are the point: a gap in a published roster is the thing nobody notices until the morning it happens. Optionally narrowed to the people on one shift template.
tasks_today_status
ChatGPTEvery task due today: fully completed, partially completed or not started, with who completed and who has not.
unfulfilled_tasks
ChatGPTEmployees with incomplete or overdue tasks on a given date, ranked by how many.
unrostered_staff
ChatGPTEmployees whose recent shifts were INVENTED by the system because they clocked in with nothing rostered. Worst first, with how many of those days were auto-closed early. The cause is a missing roster entry, never a punch problem — fix it by scheduling them, not by editing their attendance.
update_office_location
ChatGPTChanges one site — its name, address, coordinates, geofence radius, timezone, or whether it is in use. Anything left out keeps its current value. Switching it off ("is_active" no) is the safe way to retire a site: the row and everybody's assignment to it survive, and it can be switched back on. Moving the coordinates changes where people can clock in from, immediately. The answer says how many people are based here.
update_org_holiday
ChatGPTChanges one holiday by id — its name, date, working flag, description or whether it recurs. Anything left out keeps its current value. The preview shows the current row and the row it would become.
update_shift_assignment
ChatGPTEdits a single shift — its date, times, break, note, status or template. Anything left out keeps its current value. Refused if either the old date or the new one falls inside a locked payroll period, because that period has already been paid. The answer says how many attendance punches are attached to the shift, so you can see when you are editing a shift somebody already worked.
update_shift_rules
ChatGPTChanges the organisation's labour rules. Anything left out keeps its current value. These rules are ENFORCED: after this, saving a shift that breaks one is refused, and the roster you already have is re-judged against the new numbers the next time anybody edits it. Tightening a rule does not delete existing shifts, but it can make them impossible to change. The preview shows every value that would move.
update_shift_template
ChatGPTChanges one shift pattern. Anything left out keeps its current value. IMPORTANT: this does NOT change the TIMES of shifts that are already scheduled — each shift keeps the times it was created with. The answer says how many scheduled shifts are now on the old times; reapply_template_times is what moves them. Everything else on the pattern does reach those shifts at once: their name and colour on every roster, and the late grace, early-leave grace and rounding their punches are judged by. Setting active to "no" retires the pattern without touching any shift.
update_task
ChatGPTChange an existing task's name, description, priority, deadline or assignee. Its assignees see the change. An empty description clears the stored one; assign_to adds an assignee and never removes one. Requires the gpt.write scope and organisation opt-in.
upsert_accounting_reference
ChatGPTAdds a cost centre or payment method, or renames an existing one when you supply its id. Both are labels the rest of the module selects from and neither carries a balance, so this is not a money-touching change. Removing one is deactivate_accounting_record, never a delete: history references them.
upsert_org_holiday
ChatGPTAdds a holiday to the organisation calendar, or updates the one already on that date with that name. Holidays appear on every employee's calendar and the shift scheduler skips non-working ones. Replacing is a FULL overwrite: an omitted description is cleared and an omitted is_working or is_recurring resets to "no". The preview says whether this creates a new holiday or replaces an existing one, and shows both versions.
weekly_digest
ChatGPTThis week's roll-up: hours worked, tasks completed, top performer, overdue tasks and today's presence.
who_is_present
ChatGPTEmployees currently working or on break, plus who has already clocked out today.