LVNDRby Toolfora

HelpReading the change log

Reading the change log

Columns

ColumnWhat it holds
WhenRelative time (“3 min ago”). Links to the change page.
ResourceTitle on the first line, type underneath: Product, Collection, Page, Blog, Blog post, Menu, Theme or Metafield definition.
What changedCreated, Edited or Deleted. Underneath: a plain-language summary if summaries are on, otherwise the list of fields that moved.
WhoThe actor badge and a short note on how confident the attribution is.

Actor badges and tiers

Shopify tells apps who did something only for some actions. LVNDR shows how it knows, every time.

BadgeTierMeaningNote under the badge
A staff name, an app name, or ShopifyexactShopify's admin event log named this actor. Happens for creates, publishes, unpublishes, status changes and some deletes.named by Shopify
Name followed by (agent)exactThe named app looks like an AI agent, a CLI or an automation tool. Shown with a warning colour.named by Shopify
Staff via Shopify Web, or an app namechannelShopify said a staff member or an app did it, but did not render the name. Typical for deletes.Shopify names the app, not the person
Likely nameinferredNo event for this edit. The same actor created or changed this record within five minutes, so it is probably them.best guess from nearby activity
Likely bulk/appinferredNo event, but the same fields changed on five or more records within about four minutes. That is the shape of an import, bulk edit or app sync.best guess from nearby activity
Not recordedunknownNothing to go on: Shopify keeps no record of who edits a field.Shopify does not say who
Agent name followed by (agent)inferredThe change happened while an agent's announced run was open and the run collected it. The agent said it was working; Shopify did not name it.from its announced run (links to the run)
lvndr undoexactAn undo made in LVNDR, by you or by an agent through its key.made in LVNDR

A change that belongs to an agent run also carries a small “Run: agent” badge that opens the run.

The agent kind

Shopify names the API client behind a create, publish or delete. When that client's title contains words like agent, Sidekick, MCP, Claude, ChatGPT, GPT, OpenAI, Copilot, Gemini, Shopify CLI, CLI Connector, Zapier, Make.com, n8n or automation, the badge says so.

The Shopify CLI Connector App is recognised by its client id as well. Settings > “Treat these as agents” adds your own app titles and client ids. Agent actors also trip anomaly alerts sooner. See Alerts.

Why field edits say “Likely” or “Not recorded”

Shopify's admin event log records an actor for lifecycle actions only: create, publish, unpublish, status change, delete.

For a field edit (a price, a title, a tag, a metafield) it records nothing. No webhook carries an actor either.

So LVNDR infers. If a named actor touched the same record within five minutes before or after, the edit is marked “Likely” that actor. If many records changed the same way at once, it is marked “Likely bulk/app”. Otherwise it is “Not recorded”.

A guess is always labelled as a guess. LVNDR never shows an inferred name as a fact, and it never names the staff member behind a plain field edit.

Metafields, menus, themes and metafield definitions get no attribution from Shopify at all. Changes to them are inferred or unknown.

The change page

Click a row to open the change. The page shows:

  • Summary: the record type with a link to open it in the admin, when it happened, the actor badge, and the attribution note (for example “Shopify records no actor for field edits”).
  • A link to the batch when the same actor made other changes within ten minutes, and a link to the record's history.
  • Changed fields: every tracked field that moved, with its before and after value.
  • Shopify reported: the field paths Shopify's own delivery listed as added, updated or removed.
  • Matching admin events: the admin event entries that were linked to this change, if any.

The header has the undo button (“Undo this change”, “Undo: delete this record” for a create, “Restore this record” for a delete), “Put back to before this”, and “Check the store for newer edits”, which reads the record from Shopify again and logs anything that differs. Undo is hidden on a change LVNDR saw for the first time, since there is no before state to go back to.

Notes you may see

NoteMeaning
Shopify records no actor for field editsA plain edit with no nearby named activity.
Found by the daily sweep; the delivery for this change never arrivedThe daily full comparison found a difference no event had reported.
Found by the collection check; Shopify sends no event for rule, source or SEO edits on collectionsCollections are re-read every ten minutes for exactly these edits.
Found by the navigation check / theme check / definitions checkThe same periodic re-read, for menus, themes and metafield definitions.
Smart collection membership recalculated by Shopify after a product or rule changeShopify moved a product in or out of an automated collection.
Likely a bulk edit, import or app sync: … changed on N records within 4 minutesThe same fields changed on many records at once. A restore point was placed just before the first one.
Made during the agent run “…” that … announcedThe change was collected by an open agent run. Shown as the run's agent, as a guess.
Undo of an earlier change, requested in LVNDR by staff user …LVNDR itself wrote this, on request. Shown as the actor “LVNDR undo”.
Restored by undo, requested in LVNDR …A deleted record was rebuilt by LVNDR.

Summaries

When the server has an Anthropic API key, each new change gets a one-sentence summary written once and stored with it. The log shows that sentence instead of the raw field list.

Only the diff is sent to the model, and it is told to invent nothing. When summaries are on, Settings has a Summaries card that says so. Without the key, the field list is shown.

The “Who made changes” panel

The panel on the home page counts the last 30 days of changes: how many were named by Shopify, how many are a best guess, and how many are not recorded. It is a quick read on how much of your history has a known actor.