fetch
ChatGPTUse this after search to retrieve one Franlever location or AI COO activity item.
getBriefHistory
ChatGPTRead the briefs AI COO has already sent this reader over a recent window: date, daily or weekly, headline, opening sentence, each store's lever with what to do and where it went, and the report lines. Call it when the operator asks what a past brief said, what led on a given day, or how this week's brief compares to last week's. These are the documents as sent; quote them as what the brief said on that date, never as today's figures. Location ids come from list_locations.
getBusyHours
ChatGPTRead one store's busy hours exactly as the Busy hours card shows them: average sales and orders for every trading hour of every weekday, in store time, over a date range (default the last 12 weeks, at least three). Returns each weekday's busiest hours with ties marked, the single busiest slot, and weekdays the store is closed or rarely open. Call it for when a store is busiest or quietest, how Friday compares with Tuesday, or whether the rush moved over a range. Name the window it returns. Opening and closing hours are not a lever: never suggest changing them. For staffing against the rush use getHourlyCoverage; for one day use getDayDetail. Location ids come from list_locations.
getChannelBreakdown
ChatGPTRead one store's sales by channel or destination as its POS names them (dine in, drive-thru, online ordering, delivery, catering), for a recent window against the equal window before: orders, net sales, average order, share of sales, large orders (five times the store's median ticket), and which half moved — fewer orders or smaller orders. Call it whenever the operator asks why a channel is up or down, what it is costing, or how order values compare across channels. Use the driver the tool computes rather than inferring one: a channel that lost its large orders is a group-order problem, not a channel outage. Defaults to 28 days, up to 56. Location ids come from list_locations.
getDayDetail
ChatGPTRead one business day at one store from its POS: net sales, orders, average ticket, hourly labor % and labor hours, discounts, the day against a typical same weekday (median of recent ones), sales and orders by hour beside the people on the clock, sales by channel, and the largest single order's share of the day. Call it for any question about a specific day — what happened last Saturday, why yesterday was slow, how Tuesday compared — rather than asking the operator for a POS export. Use last-weekday for 'last Saturday' so the store's calendar decides the date. Quote the comparison basis the tool gives, and when one order carried the day, say so. Location ids come from list_locations.
getFranchiseDisclosures
ChatGPTRead franchise disclosure filings (FDDs) as franchisors filed them with state regulators: Item 19 sales and cost figures, Item 20 outlet counts, unit growth and franchisee turnover, and Items 5-7 franchise fee, initial investment, royalty and ad fund. Use scope 'brand' for this workspace's own filing — royalty, fees, what we disclosed for unit volume — and scope 'industry' to place a brand against its segment. These are annual filings, not live data: name the issue year, say when a filing is stale, carry the sampling caveats, and never build a store-level action on them. Location ids come from list_locations.
getHourlyCoverage
ChatGPTRead one store's orders and sales against the people on the clock, hour by hour in store time, averaged per trading day over the last few weeks and split by weekday group (Mon–Thu, Fri–Sat, Sun) or by single weekday. Each group marks its peak hour and the orders per person it runs there, and flags hours carrying more people than that pace would need. Call it for any question about labor or staffing by hour or day, when the rush is, or whether a shift is over- or under-staffed — never send the operator to pull a labor report. People counts include salaried managers. A staffing gap is people, not dollars: never price it. Defaults to 4 weeks, up to 8. Location ids come from list_locations.
getLeverHistory
ChatGPTRead what AI COO has raised at stores over a recent window, and what came back: each lever's title and detail, the version the GM was sent, when it was first raised and last confirmed, how many briefs carried it, and the GM's answer — done, later, or wrong — with any note they left. Call it when the operator asks what was flagged before, whether something has come up already, what a GM said, or how a store has responded. A lever with no answer says nothing about whether the work happened, because Franlever sees only replies: never call one overdue, late, outstanding, unanswered, ignored or slipping, and never count open levers as a failing. 'wrong' is a GM saying the call was mistaken — quote their note and treat it as information about the finding. Location ids come from list_locations.
getMarketContext
ChatGPTRead the outside-world context for a store: actionable weather for today and tomorrow, county wages, menu-price inflation, ingredient cost pressure, and product recalls. These are the denominators — a store up 3.1% in a market up 3.3% went backwards — so call it before calling a move a win or a loss, and whenever weather, wages, or food costs could explain what the operator is asking about. State the effective date and never present a stale market figure as current. Location ids come from list_locations.
getMetricCorrelations
ChatGPTRead measured relationships between a store's own metrics over its last 60 days — labor versus sales, waste versus sales, reviews leading sales, web demand versus sales. Call it when the operator asks whether one thing is driving another, and cite the strength and window rather than asserting cause. A correlation is evidence for a hypothesis, never proof of one. Location ids come from list_locations.
getProductMix
ChatGPTRead exact item-level POS sales for one store and a requested business date or date range. Use relative-day for phrases such as today, yesterday, or two days ago so the date is resolved in the store's timezone. Use this whenever the requested window differs from the rolling productMix summary already in context. 'Most popular' or 'best selling' means units unless the operator explicitly asks for revenue. Location ids come from list_locations.
getStoreSignals
ChatGPTRead what a store's own POS data says, computed: ticket versus traffic, weekday range breaches, record days, streaks, month-to-date pace, closed days, sales per labor hour, order concentration, channel mix and dormant channels, cash share, speed of service, discount spread, item attachment, and overtime concentration. This is the exact evidence AI COO's scan builds its actions on, so call it whenever the operator asks why a metric moved, what is going on at a store, or about a finding in their brief. Each signal says whether it is actionable or informational and whether it is good, bad, or neither — do not turn an informational signal into a task. A signal carrying weeklyDollars is one Franlever could put a figure on: it is what closing the gap to this store's own recent normal is worth per week, so quote it as recovering that gap and never as new revenue or as a target. A signal without one is not worth nothing — it is one Franlever will not guess at, so say what moved without inventing a number. Reads up to 12 stores per call. Location ids come from list_locations.
getWeekdayPatterns
ChatGPTRead one store's normal week from its POS: what each weekday usually does, each day's busiest hours in store time, and what sells on each day. Call it for questions about a named day or hour at a store — which day is slowest, what sells on Sundays, when the lunch rush peaks. Name the window it returns, and use the rankings it marks rather than comparing the figures yourself; a day marked level with others is not a ranking to pick from. Location ids come from list_locations.
get_location_performance
ChatGPTUse this when you need sales, labor, orders, and average-ticket history for one authorized store.
get_portfolio_snapshot
ChatGPTUse this when you need a current, deterministic summary of sales, labor, reviews, waste, and stores needing attention.
list_ai_coo_activity
ChatGPTUse this when you need recent AI COO observations and levers already delivered for authorized stores.
list_locations
ChatGPTUse this when you need the stores, current metrics, or current lever in one Franlever workspace.
list_workspaces
ChatGPTUse this when you need to see which Franlever workspaces the user authorized for this assistant.
search
ChatGPTUse this when you need to find Franlever locations or recent AI COO activity by name, place, metric, or lever.
searchReviews
ChatGPTSearch customer reviews for stores over a recent window, filtered by rating and by a word or phrase in the text — 'cold fries', 'wait', 'rude'. Returns up to 25, newest first, with store, rating, date and text. Call it for questions about what customers are saying, recurring complaints, or low ratings at a store; the handful of recent reviews already in context is not a search. Quote reviews rather than paraphrasing them into a pattern the results do not show, and say how many you read. Location ids come from list_locations.
showMetricComparison
ChatGPTRender an inline bar chart comparing one metric (sales, labor %, avg ticket, or orders) across stores over a window, ranked best-first with the brand median marked. Call this when you compare stores on a metric or point out which stores lead or lag. Needs at least two stores with data for the window; for a single store use showMetricTrend instead. Real POS data — do not restate every bar in prose. Location ids come from list_locations. Over MCP this returns the chart's data points rather than drawing it.
showMetricTrend
ChatGPTRender an inline line chart of one store's daily metric (sales, labor %, avg ticket, or orders) versus the brand median. Call this whenever you describe how a single store's metric is trending over time. One chart per store and metric: never call it twice for the same question. Numbers come from real POS data — do not invent values; keep your prose about the chart to one short takeaway. Location ids come from list_locations. Over MCP this returns the chart's data points rather than drawing it.
showPortfolioScatter
ChatGPTRender an inline scatter plot of revenue versus labor % across every store over a window, so outliers stand out. Call this when discussing the portfolio's overall efficiency or which stores are high-revenue but high-labor. Real POS data. Location ids come from list_locations. Over MCP this returns the chart's data points rather than drawing it.
fetch
ClaudegetBriefHistory
ClaudegetBusyHours
ClaudegetChannelBreakdown
ClaudegetDayDetail
ClaudegetFranchiseDisclosures
ClaudegetHourlyCoverage
ClaudegetLeverHistory
ClaudegetMarketContext
ClaudegetMetricCorrelations
ClaudegetProductMix
ClaudegetStoreSignals
ClaudegetWeekdayPatterns
Claudeget_location_performance
Claudeget_portfolio_snapshot
Claudelist_ai_coo_activity
Claudelist_locations
Claudelist_workspaces
Claudesearch
ClaudesearchReviews
ClaudeshowMetricComparison
ClaudeshowMetricTrend
ClaudeshowPortfolioScatter
Claude