Management · your online store, run for you
Your store, run end to end.
The storefront, the catalog, the orders, the restocking and the people who ask — all of it run and watched, so the store keeps selling while you do something else. Infrastructure built it; Management keeps it selling.
The whole store, as four jobs that never stop — not a tool you have to drive yourself.
Fast, up and selling — pages, checkout and payments watched around the clock. Slow or down is caught before a customer feels it.
Listings kept current, prices right, every order seen from click to delivery. Nothing sits unfulfilled and nothing oversells.
Restocking proposed before you run out, suppliers chased, costs watched. The buying side runs as tightly as the selling side.
Customer questions answered in your voice, returns handled, the awkward cases escalated to a human. Your buyers are never left waiting.
The difference is not a flashy dashboard. It is the watch — and nothing acts on your store on its own.
How the store is held — not invented growth figures.
The agents are not the show; they are one more managed layer under the same watch.
Here is the real shape of it.
The code stays in English — the universal language of engineering.
# one store under watch — what raises a hand, and what it does on its own (nothing) POST https://api.matchtopalgorithm.ai/v1/operate/stores Authorization: MT-••••-••••-•••• — one key, one store Content-Type: application/json { "store": "shop.yourcompany.com", "watch": ["uptime", "checkout", "stock", "price_drift", "orders", "messages"], "on_breach": { "alert": ["telegram", "email"], "act": "propose_only" }, "window": "24/5", "clock": "Europe/Zurich" — the sheet arrives in your hour } 202 Accepted { "store_id": "s_7f3a91c4", "watch": "live", "autopilot": "off — a human signs every action" }
"""MatchTop · a store agent — it proposes, it never ships on its own. Standard library only. The agent raises a hand; a person decides.""" import smtplib, json from dataclasses import dataclass @dataclass class Proposal: kind: str # "restock" | "refund" | "reprice" detail: dict signed: bool = False # stays False until a human says yes def watch_stock(store, floor=3): """Low stock is not an order. It is a proposal that waits for a signature.""" for item in store.low_stock(below=floor): yield Proposal("restock", {"sku": item.sku, "qty": item.reorder_qty}) def apply(p: Proposal): if not p.signed: raise PermissionError("no action ships without a human signature") store.execute(p) # only reachable after sign-off
"""MatchTop · the close — sales out, a sheet your accountant can open. No one sits to compile it; it is a read of what is already recorded.""" import csv, sys def close(period, sales) -> str: """Month or quarter: one clean export, in your currency, nothing invented.""" path = f"close-{period}.csv" with open(path, "w", newline="") as f: w = csv.writer(f) w.writerow(["date", "order", "item", "qty", "net", "tax", "gross"]) for s in sales: w.writerow([s.date, s.order, s.item, s.qty, s.net, s.tax, s.gross]) return path # handed to the accountant, not to a black box
Why a run store beats a store you must run yourself.
A store tool cannot run your store — it hands the running back to you. A faceless service cannot give you the one engineer who watches it. And an agent that acts on its own cannot be trusted with your money. That is why there is no price list here: we look at what the store must do, we measure it, and we quote what we measured.
Before a line is written, how your store and its data are kept. This part is not negotiable.
An NDA is the default here, not an upsell. What you share is read by one engineer and no one else — nothing is sold, shared, or passed on.
Store keys, customer and sales data stay in your house, under least privilege. We keep no copy we do not need, and nothing leaves your reach without your word.
Every agent action is logged and signed — you can read who did what, and when. No black box moving your money while you look away.
Two lines — what the store sells, and what you want off your hands. We read them ourselves and we answer ourselves.
It is already on a screen in Zürich — not in a queue, not in a ticket. The engineer who would run this reads it, and writes back himself.
·