Sheet 01 · The record

Nothing ison fire yet.That's the job.

The SKU that runs out in nine days. The PO two weeks late. The return rate that tripled on one variant. Mnemos keeps the record and runs the same twelve checks every week — so the quiet problems get caught while they're still quiet.

Install on ShopifyStart free15-minute census · read-only · 14-day trial
Weekly review · run 38live
01Stockout before restockSKU-4492 · 11 d cover, PO lands d16FLAGGED
02Reorder now3 SKUs below velocity floorFLAGGED
03Late purchase ordersPO-0917 · 14 d past expectedFLAGGED
04Lost sales to stockouts$2,140 est. last weekCLEAR
05Bundle candidates2 pairs, 71% co-purchaseCLEAR
06Return-rate outliersVariant "Slate / L" · 3.1× baselineFLAGGED
07Slow movers18 SKUs, 0 units / 60 dCLEAR
08Stale listingsNoneCLEAR
09Margin-eating discountsCode SPRING20 · −4.2 ptsFLAGGED
10Velocity shifts+38% on "Field jacket"CLEAR
11Channel-mix shiftsWholesale share 22 → 27%CLEAR
12Agent insightShopify + NetSuite disagree on 2 countsTHINKING
01Stockout before restockSKU-4492 · 11 d cover, PO lands d16FLAGGED
02Reorder now3 SKUs below velocity floorFLAGGED
03Late purchase ordersPO-0917 · 14 d past expectedFLAGGED
04Lost sales to stockouts$2,140 est. last weekCLEAR
05Bundle candidates2 pairs, 71% co-purchaseCLEAR
06Return-rate outliersVariant "Slate / L" · 3.1× baselineFLAGGED
07Slow movers18 SKUs, 0 units / 60 dCLEAR
08Stale listingsNoneCLEAR
09Margin-eating discountsCode SPRING20 · −4.2 ptsFLAGGED
10Velocity shifts+38% on "Field jacket"CLEAR
11Channel-mix shiftsWholesale share 22 → 27%CLEAR
12Agent insightShopify + NetSuite disagree on 2 countsTHINKING
Sheet 02 · Fifteen minutes to a working picture

It reads the whole operation before it says anything.

No modelling exercise, no data team. Install it read-only and it walks the catalogue, the locations, the open POs and the last ninety days of orders — then tells you what it found.

Mnemos — census of the operation
MN-01
Sheet 02 · Fifteen minutes to a working picture
01Catalogue · SKUs and variants4,218
02Channels + locations6 · 3
03Orders on record, trailing 90 d12,640
04Open purchase orders14
05Time to first answer14 min
06Weekly checks, deterministic11 + 1
Census in progress — figures update as the read completes.reading
01
Install read-only

One grant from the Shopify App Store. No engineering, no warehouse, no data team — and no write scopes to revoke later.

02
A 15-minute census

Catalogue, channels, locations, inventory, orders, fulfillments, POs, returns — read end to end, then reported back as a count.

03
Ask it something hard

Plain language in, an answer out, with the rows it read attached. If the data will not support an answer, it says so.

04
Read Monday's review

The same twelve checks against current data every week, with what changed since the last run.

Sheet 03 · Four verbs

Answer. Watch. Remember. Review.

Mnemos is not a dashboard you have to interpret. It is an employee who has read the record and will show you the rows it read.

YouWhich SKUs run out before their inbound PO lands?
MnemosFour. SKU-4492 is tightest — 11 days of cover at its current rate, and the PO covering it is due in 19. The other three are borderline; each is listed with the assumption behind its date. 412 rows attached ↗
Sheet 04 · The weekly review

Twelve checks. The same twelve, every week.

Eleven are deterministic rules over your own data — same inputs, same output, every time. The twelfth is whatever the agent noticed that the other eleven would have missed. Nothing new is shipped on a whim; every check has a name and a row.

01Rule
Stockout before restock

SKUs whose cover runs out before the purchase order that would refill them arrives.

stockout_before_restock
02Rule
Reorder now

What you're low on this week, given lead time, current velocity, and what's already inbound.

reorder_now
03Rule
Late purchase orders

POs past their expected date, ranked by the demand each one is holding up.

late_purchase_orders
04Rule
Lost sales to stockouts

Demand you couldn't fill last week because something was out of stock, priced at your own margin.

lost_sales_to_stockouts
05Rule
Bundle candidates

Products bought together often enough, and at enough margin, to be worth selling together.

bundle_candidates
06Rule
Return-rate outliers

Variants returning well above their own baseline, with when the shift started.

return_rate_outliers
07Rule
Slow movers

Inventory that has not moved in weeks, and what holding it is costing you.

slow_movers
08Rule
Stale listings

Live products with no sales and no changes — the quiet tail of the catalogue.

stale_listings
09Rule
Margin-eating discounts

Codes and price points moving volume below the margin you set.

margin_eating_discounts
10Rule
Velocity shifts

Products accelerating or falling away against their own trend, not against an average.

velocity_shifts
11Rule
Channel-mix shifts

Where demand moved between channels and locations, and what moved with it.

channel_mix_shifts
12Agent
Agent insight

What the agent noticed this week that none of the eleven checks are built to see.

agent_insight
Every check ships with its name and the query behind it. Nothing changes without a revision.Rev. 38 · run every Monday 06:00 local
Sheet 05 · Integrations

Shopify first, then everywhere else the truth lives.

The census starts with your store. Every other connector adds a column to the same picture — never a second picture to reconcile.

ShopifyWhere Mnemos starts

Install from the App Store and the census begins immediately — no keys to paste, no pipeline to build. Mnemos requests read scopes only: it can see your operation and it cannot change it.

ordersproductsinventorylocationsfulfillmentsreturnscustomersshipping
What Mnemos reads, exactly ↗
NNetSuitePurchase orders, items, receipts
QuickBooksCosts, invoices, margin
ZendeskTickets against orders
GorgiasTickets against orders
SnowflakeYour own warehouse tables
LLogiwaWarehouse and fulfillment
NotionHow the team says it works
ConfluenceRunbooks and policy
Running something else? Connectors are declarative documents, so a custom one is a config file against your API — not a project.
Sheet 06 · Ask from Claude or Cursor

Or ask from Claude. Or Cursor.

Every workspace gets its own MCP server. Point your editor or assistant at it and your operation is just there, in the tool you already had open.

  • 01One server per workspace, scoped to that workspace's data
  • 02The same read-only access the web app has — nothing more
  • 03Answers arrive with their evidence there too
claude_desktop_config.json
mcp
read-only
{
  "mcpServers": {
    "mnemos": {
      "url": "https://api.mnemos.ai/mcp/acme-supply",
      "headers": {
        "Authorization": "Bearer mn_live_…"
      }
    }
  }
}
Sheet 07 · Pricing

A base plan and a credit meter.

1 credit ≈ 1 answer. Seats and connectors are unlimited on every tier, and the first 14 days are free — charging per seat would only teach your team not to ask.

Pricing — a base plan and a credit meter
USD / mo
14-day trial
No.TierCredits / moPrice
01FreeRun the census, read the first weekly review.0$0
02StarterOne store, one operator, questions all week.1,000$499
03GrowthMost pick thisA team asking through the week, across systems.3,000$699
04ScaleSeveral stores, several channels, heavy use.10,000$999
Billed through Shopify when installed from the App Store, otherwise by card.Credit packs, and what a credit is ↗
Sheet 08 · Close