Skip to content

Epinium MCP ​

Access: Settings → MCP

The Epinium MCP lets an AI assistant query your Epinium data directly, with no exports and no code. You can ask it in plain language which campaigns have the worst ACOS or which searches are eating your budget, and it answers with your real figures.

It runs on the Model Context Protocol, an open standard for connecting assistants to data sources. Under the hood it uses the Epinium API, so it sees exactly the same data and honours the same permissions.

You decide whether it can write

By default the assistant only queries. The tools that change anything sit behind write permissions that you tick when you create the key -workflows:write for your automations, campaigns:write for your campaigns on Amazon, tasks:write for your tasks, catalog:write for your keyword clusters, your prompts, your brand identity, the changes to your Amazon listings and your new product listings, catalog:publish for it to publish on Amazon the pending items you confirm, and costs:write for your product cost, its purchases and your own expenses-: withhold them and those tools do not exist for the assistant. In your Shopify and WooCommerce stores, the same tool it queries with is the one that also writes, and it only does so if the key carries platform:write: without it, any attempt to change something is rejected before it reaches the store. And writing is for the Business and Master plans: on Free and Guru the assistant queries, but cannot change anything.

The server address ​

https://mcp.epinium.com/mcp

You will also find it in Settings → MCP, with a button to copy it.

How to connect it ​

There are two ways to give the assistant access, and which one applies depends on the client you use:

  • Browser authorisation: you paste the server address, the assistant takes you to Epinium, you approve and that is it. There is no key to copy, and you can withdraw access whenever you like from Settings → MCP.
  • API key: you create a key in Settings → API Keys and paste it into the client's configuration. This is the route for Claude Code and for your own integrations.

Claude (web and desktop app) ​

  1. Open Claude's connector settings and add a custom connector
  2. Paste https://mcp.epinium.com/mcp as the server address
  3. Claude will take you to Epinium to authorise access: review the permissions it asks for and approve them
  4. Check that it works by asking it to use the epinium_whoami tool

Custom connectors require a paid Claude plan.

Claude Code ​

Claude Code is configured from the terminal and uses an API key rather than browser authorisation:

bash
claude mcp add --transport http epinium https://mcp.epinium.com/mcp \
  --header "Authorization: Bearer TU_API_KEY"

Replace TU_API_KEY with a key created in Settings → API Keys, with whichever permissions you want to give the assistant.

ChatGPT ​

ChatGPT's MCP connectors live in developer mode (Settings → Security and login), and they are only available on paid plans and from the web version.

  1. Turn on developer mode under Settings → Security and login
  2. In the connector settings, add a new connector
  3. Paste https://mcp.epinium.com/mcp as the server address and leave the registration method as it comes: the connector signs itself up
  4. ChatGPT will take you to Epinium to authorise access: review the permissions it asks for and approve them
  5. Check that it works by asking it to use the epinium_whoami tool

What you have connected ​

Settings → MCP lists the assistants you have authorised, with which permissions, who approved them and when. From there you can withdraw access from any of them: it immediately stops being able to read your data, and you can authorise it again later if you want.

API keys are managed separately, in Settings → API Keys.

What tools it includes ​

The assistant sees one tool per resource your token reaches. The names are descriptive, so you can just ask for what you want in plain language.

GroupToolsPermission
CatalogProducts, their Seller, Vendor and Advertising views, the category of their rank and, for Vendor, their reviews, brands, countries, clusters, custom prompts and brand identitycatalog:read
Catalog performanceSeller and Vendor product metrics: sales, traffic, Amazon fees and margin, returns, stock, days of coverage and the Vendor sell-incatalog:read
P&L with product costYour account's waterfall down to net profit, margin by product, by sales channel, cross advertising and data freshness, with product cost and own expenses if your plan includes itcatalog:read
Own expensesYour own expenses (staff, agency, 3PL, software…) and their categoriescatalog:read
Advertising structureCampaigns, ad groups, targets, product ads and search termscampaigns:read
Advertising performanceMetrics for the five levels abovecampaigns:read
PortfoliosYour portfolios and the budget cap each one puts on its campaignscampaigns:read
Advertising writesCreate and change campaigns, portfolios, ad groups, targets and product ads on Amazoncampaigns:write
Optimization configWhat the optimizer chases on a campaign: objective, bid limits, budget and which workflows process itcampaigns:write
AutomationYour workflows, their configuration, which campaigns they reach and how those campaigns performworkflows:read
Writing automationCreating, copying, editing, activating, pausing, deleting and restoring workflows, running or simulating them, and the forecast of what it would costworkflows:write
Automation changesWhat a workflow changed on a campaign, with the net effect per entitycampaigns:read
Change historyEverything Epinium has changed on Amazon, whatever set it off, including what Amazon rejectedcampaigns:read
ConnectionsYour connections: Amazon, Shopify and WooCommerceconnections:read
Connected storesepinium_platform_docs and epinium_platform_request: look up what is inside your Shopify and WooCommerce stores and, with platform:write, create, edit and delete products and content with the same tool. Writing is direct, with no task, and some deletions cannot be undoneplatform:read, and also platform:write to write
Store changesepinium_list_platform_changes: everything written to your stores in the last 12 months, including what the store rejected. It costs 1 credit per row returnedplatform:read
TasksEpinium tasks and their itemstasks:read
Writing tasksCreating tasks with suggestions, approving or rejecting them, closing them and applying what was approved, and asking Epinium's AI to rewrite up to 20 listings (epinium_optimize_listings)tasks:write
Writing clustersCreating keyword clusters, adding or removing their keywords, products and prompts, and deleting themcatalog:write
Writing promptsCreating and editing the custom prompts the AI follows when it writes your listingscatalog:write
Writing the brand identityepinium_create_brand_identity and epinium_update_brand_identity: creating and editing your brand identity, with a profile per competitorcatalog:write
Listing changesPreparing changes to any attribute of your listings, from what you tell it in the chat or from a spreadsheet: finding their products, matching each piece of data to an Amazon attribute and validating the rows without saving anythingcatalog:read
Writing listingsSaving the rows of a spreadsheet, or the ones you dictate in the chat, as changes pending to publishcatalog:write
New listings on Amazonepinium_search_amazon_catalog, epinium_search_amazon_product_types, epinium_list_amazon_shipping_templates, epinium_validate_listing_draft and epinium_list_listing_drafts to find the product, choose its type, validate the new listing and follow it up; epinium_create_listing_draft and epinium_discard_listing_draft to save it as pending or discard it. Offers, new products and families, on Seller and on Vendorcatalog:read, and also catalog:write to save and discard
Publishing to Amazonepinium_review_and_publish and epinium_list_pending_listing_changes to see what is pending to publish, changes and new listings; epinium_publish_listing_changes to publish what you confirm, the same way as the app's Publish buttoncatalog:read, and also catalog:write and catalog:publish to publish
Writing product costepinium_set_product_cost, epinium_preview_product_costs and epinium_import_product_costs: creating or updating the unit cost and shipping cost of one product, or checking and importing those of many at once from a filecosts:write
Purchasesepinium_list_purchases, epinium_get_purchase, epinium_list_product_purchases and epinium_get_product_stock_group to read them; epinium_record_purchase, epinium_update_purchase and epinium_delete_purchase to record, correct and delete themcatalog:read to read, costs:write to write
Writing own expensesCreating expenses one by one or in a batch, changing their concept, category or amount, ending and deleting them, and managing your categoriescosts:write
SkillsThe CMO Brain: Epinium's 101 marketing skillsskills:read
Discoveryepinium_find_data: where each piece of data lives. The assistant asks it for a name, an acronym or a synonym (BSR, ACoS, TACoS, Buy Box, margin) and it answers which tool and which field hold it. It does not read your data or spend creditsany of the above
Accountepinium_whoami and the list of accessible accountsalways

Only the tools your token can use appear. If you create a key with campaigns:read alone, the assistant will not see the catalog or task tools: they do not fail, they simply do not exist for that key.

CMO Brain: Epinium's marketing brain ​

Beyond your data, the MCP serves the CMO Brain: 101 skills with the methods a CMO would apply, written so an assistant can read and follow them. They cover everything from advertising and SEO to pricing, retention, content and marketplaces, and the assistant consults them before advising you.

The difference shows: without them, an assistant answers with whatever it read online; with them, it follows the same criteria we work by and applies it to your figures. Ask it to review a campaign with a poor ACOS and it does not improvise generic advice: it loads the relevant skill and follows it.

They require the skills:read permission, which is ticked by default when you create a key.

Account totals, by period and by product family ​

The assistant answers big-picture questions with a single query: "how much did I sell in September?", "how has my advertising spend gone month by month this year?", "how are the products in my coffee machine cluster selling?". The metrics tools add up by account, by country or by connection on their own, split the range into weeks, months, quarters or years, and can be narrowed down to one or more of your product clusters.

That way it does not have to walk your catalogue product by product and add it up, which is slow and expensive: a month's total for a large account costs about 150 credits instead of more than 17,000.

Four things worth knowing when reading its answers:

  • Each currency comes separately. If you sell on marketplaces with different currencies, the assistant receives one total per currency: amounts are never converted.
  • Stock is only added up by country. FBA's pan-European inventory shows up in full in every marketplace, so an account-wide stock total built by adding up countries would not be a real figure. The assistant does not receive it; if you ask for it, it gives it to you country by country.
  • By cluster, each product counts in all of its clusters. Ask how much each cluster sold and it gets the answer from a single query; a product in two clusters adds up in both, so the cluster totals do not make up the account total.
  • If it starts adding up product by product, it is told not to. When the assistant asks for an ungrouped list of metrics and more pages follow, the response itself tells it to group instead of paging on.

The P&L with product cost ​

With the catalog:read permission, the assistant can read your full profit and loss statement: the same one you see in Profitability → P&L, spread across ten tools -epinium_get_pnl_cascade, epinium_list_pnl_products, epinium_get_pnl_product, epinium_list_pnl_clusters, epinium_list_pnl_channels, epinium_list_pnl_ads_matrix, epinium_get_pnl_freshness, epinium_list_pnl_margin_leaks, epinium_get_pnl_category_medians and epinium_list_pnl_inventory-, with the same calculation and the same figures.

What it sees depends on your plan. On Free and Guru the assistant reaches margin_before_ads -what is left after Amazon, before advertising and before product cost- and it will not call that "profit", because it is not. On Business and Master it reaches the final number: contribution margin, with every product-cost line and its coverage, and your own expenses down to net profit. Fixed expenses belong to the whole account: if you have any and filter by connection, country or cluster, or if your token does not reach every sales connection, the waterfall ends at contribution margin and the assistant knows why. The includes_product_cost field on every response tells it which of the two it is looking at, so it does not have to guess from whether the figures match what you expected. The two margin-leak tools are Business and Master only: on Free and Guru they return a 403, and the assistant will tell you it is on the Business and Master plans. The same goes for epinium_list_pnl_inventory, the capital tied up in Amazon's warehouses, as in Inventory. Their responses carry no includes_product_cost, because they always include cost. If you want to act on a leak, the assistant offers to create a task with those products (with the tasks:write permission), and only creates it once you say yes.

Five things worth knowing:

  • It is the only part of the MCP that converts currency. Every other resource responds in the currency each amount actually occurred in; here the assistant asks for a currency and everything is converted to it, at the exchange rate of the day each amount occurred.
  • breakeven_acos, in Seller metrics, is the ACOS ceiling you can afford to spend on advertising before losing money on that product, calculated with your real margin and your real cost -never with an assumed fee percentage. It comes back null -never a made-up number- if the product does not have a full cost, sold nothing in the window, has its cost recorded in a currency other than its own marketplace's (this figure does not convert currencies), or your plan does not include product cost (Free and Guru).
  • Product cost is not deleted when you downgrade your plan. If you move to Guru, the cost you already had loaded stays there: it just stops being served, and it comes back as soon as you upgrade again.
  • It can be narrowed down to your product clusters, and compare the clusters with each other. The cascade, the per-product table, the per-cluster table and the freshness accept one or more clusters. In the per-cluster table a product that is in two clusters counts in both. Advertising then only counts the spend of those products: Sponsored Brands and campaign spend that does not belong to a specific product are left out, because they belong to no cluster.
  • It also sees the cost of each product in the catalog. When listing your products, the assistant sees the product and shipping cost that applies today, and can keep the ones without a cost or the ones without a cost that sold in the last 30 days, to tell you which to fill in first. It is the same as what you see in Costs and expenses.

Asking it what your automation has done ​

With the workflows:read permission the assistant can audit your workflows, and it is one of the things it does best: these are cross-referencing questions, the kind that take an afternoon by hand.

You can ask it three different things:

  • How a workflow is configured — its schedule, its diagram and every variable with the value it actually carries, flagging the ones you changed against the ones still on the template default. It is the quick way to review a setting without opening the editor.
  • Which campaigns it actually reaches — which is not the same as the campaigns linked to it. A workflow can be scheduled every night and process none at all; this is the question that surfaces it, and when the set comes back empty the assistant tells you the reason.
  • What it changed on a campaign — the bid each target started from against the one it has now, how many times it moved along the way, and the reason for the last change.

What really pays off is pairing that with the metrics for the same period: "what has the workflow changed on this campaign this month and how did the ACOS go?" is a single question, and the assistant holds both halves. That is where it can tell you whether the changes are achieving anything or just moving bids around.

Two things worth keeping in mind when reading its answer:

  • Changes come as a net, not as a list. A target raised and cut back on the same night shows up as no net change, two moves. That is deliberate: a flat decision list hides exactly that.
  • It never sees simulations. Only changes that actually reached Amazon, and only from the last 90 days, which is how long they are kept.

Running a workflow, or simulating it first ​

With the workflows:write permission the assistant does not only read your automations: it can run them too. That helps when you do not want to wait for the scheduled time —you have just changed a setting and want to see the effect— and when you want to check a new workflow before letting it loose.

Two things worth asking for before running it for real:

  • What it would cost — the assistant can give you the forecast without running anything. It pays off on any workflow that walks your catalogue or your campaigns: the cost grows with the number of items it processes, so a twelve-step diagram can be expensive on a large account. It is a ceiling, not a bill; and if a step cannot be forecast, it comes back with zero items and the assistant points it out.
  • That it simulates first — a simulation walks the whole diagram and writes nothing: not to Amazon, not to your data. It tells you what it would have done, without doing it.

A paused workflow can be simulated too, and that is precisely why simulation exists. Checking an automation before switching it on should not force you to switch it on first, which is the risky step. Running it for real does require it to be active, exactly as in the app.

Two things worth keeping in mind:

  • It does not wait for the run to finish. Running returns the identifier of the execution, not its result: a pass over dozens of campaigns takes far longer than an assistant reply. The result is read afterwards, and the assistant knows how.
  • Running for real moves bids and budgets on Amazon. Simulating and running are two different requests and the assistant has to say which one it is making. If it does not, ask before you confirm.

Creating and editing automations from the assistant ​

The same workflows:write permission lets the assistant build and change your automations, not just run them. It can create one from scratch, copy one you already have, start from a template in Epinium's catalogue, change its diagram or its settings, give it a schedule, turn it on, pause it, delete it and go back to an earlier version.

Everything it creates is born paused, by any of the three routes —creating, copying or instantiating a template—. Turning it on is always a separate step you have to ask for: from then on its schedule fires for real every night.

Four things worth knowing:

  • To edit a scheduled workflow it has to be paused first. The assistant will tell you: pause, make the change, check the result and turn it back on. Pausing, on the other hand, is always accepted —it can only make fewer things happen on Amazon, never more—.
  • Copying beats rebuilding. If you want the same logic with different values, ask it to copy the automation and change only what changes. That is cheaper and more reliable than rebuilding the diagram step by step. The copy is born paused and with no campaigns attached: attaching them is done in Epinium.
  • Deleting is the only thing it will ask you to confirm. It is announced to the assistant as a destructive operation, so your AI client will ask first. The rest deliberately do not: if everything asked for confirmation, you would end up accepting without reading. What really protects you is that nothing is born switched on and that simulating and running are two different requests.
  • An invalid diagram is not saved. It is checked before writing, so an edit cannot leave your automation half-done: either the change goes in whole, or it tells you what is wrong and touches nothing.

Every write is recorded: what changed, when and with which credential. Deletion too, which leaves a trace even though the automation no longer exists, and runs, which record who asked for them. In each workflow's version history you can tell what an assistant did from what a person did in the app, and you can go back to any saved version.

If you would rather see it or touch it by hand, all of this lives in Processes → Workflows.

If the automation writes to your store on Shopify or WooCommerce, the credential also needs platform:write to save it, copy it, turn it on or run it for real. Simulating it, pausing it or renaming it doesn't. More in Reading and writing to your store.

Interactive tables ​

When you ask it for a list — products, campaigns, search terms — the assistant can show it as a table inside the conversation instead of as text.

With that table you can:

  • Load more rows without spending assistant turns. The table fetches the next pages on its own, so working through a long catalogue does not consume conversation.
  • Open a row to see its full detail, including the fields that do not fit in the table.
  • Select several rows and hand them to the assistant with the Use N in chat button. From then on it reasons about those only: pick six products out of two hundred and ask about them, without copying a single ASIN.

When the assistant can write ​

With the campaigns:write permission the assistant stops just looking: it creates campaigns and moves bids, budgets and states really, on Amazon. Grant it deliberately, and only to the key that needs it.

What the assistant itself knows, because the tools tell it:

  • What it creates is born paused unless asked otherwise, and it has to warn you before enabling anything: an enabled campaign starts spending as soon as Amazon accepts it.
  • It can archive, and it will ask you first. Archiving is the only irreversible operation, so it has its own tool and is announced to the assistant as destructive: your AI client will ask you to confirm before running it. If you only want to stop the spend, ask it to pause instead.
  • To move many bids it uses the batch tool, not the single one in a loop. One batch takes up to 1000 entities, each with its own value.
  • A batch can come back partial, and that is the normal outcome. The assistant gets one row per entity plus a count of what reached Amazon; if it does not read that count, it will assume everything went in. If it tells you a change was applied, you can ask for the breakdown.
  • It also manages portfolios. It can create them, change their name, state and budget cap, and move Sponsored Products campaigns into or out of them. The cap limits what all the portfolio's campaigns spend together, so changing it moves the spend of all of them at once.
  • It also writes to Sponsored Brands and Sponsored Display. It can change and archive their campaigns, ad groups, targets and product ads, and create their targets and negatives, with the types the ad group's ad product accepts. What it cannot do yet is create SB or SD campaigns. And for AUDIENCE, CONTENT_CATEGORY and LOCATION targets it needs the Amazon id, which you have to give it from the Amazon Ads console.

Everything it writes is recorded with its origin: in the change history it shows up as mcp, with the credential that did it. That is how you tell what an assistant did from what a person did in the application.

Configuring a campaign's optimization ​

With the same campaigns:write permission, the assistant can also read and change the configuration Epinium's optimizer uses on a campaign: its objective, bid limits, monthly budget, harvesting overrides and which workflows process it.

It is not the same as writing to Amazon, and it is worth being clear about: this configuration stays inside Epinium - the optimizer consumes it on its nightly run - so the tool does not ask for an idempotency key, nor does it leave a row in Amazon's change history, because it never reaches it.

You can ask it things like "set this campaign's target to ACOS 15%", "raise the bid ceiling to $0.80" or "take this campaign off workflow X", and the assistant translates that into the matching configuration. Before touching anything, it can also tell you the current values.

What is really worth confirming before accepting:

  • Assigning a workflow to the campaign is what puts it under the optimizer - not aiEnabled, which is only the account holder's consent and travels to Amazon as a tag, nor any other setting in this block. If the assistant is about to link the campaign to a workflow, confirm it first: from that night on, that workflow moves real bids on Amazon against it. A workflow with auto-selection can also pick it up on its own unless the campaign has declined that selection, and the exclusion list is the only thing that reliably takes it out.
  • A batch change is not all-or-nothing. If you ask it to apply the same setting to several campaigns, it gets one result row per campaign - each one validated against its own stored state - so one can fail for a specific reason while the rest go through. Ask it for the breakdown if something does not add up.

Proposing changes as tasks ​

With the tasks:write permission the assistant can leave you its proposals as a task instead of applying them itself: it analyses, creates the task with one suggestion per change, and you review it in Processes → Tasks or settle it from the conversation. It is the way to let an assistant work on your account without anything reaching Amazon before you have seen it.

Four things worth knowing:

  • Approving and applying are two steps. Approving a suggestion only records your decision; then you ask it to apply the task. Campaign changes reach Amazon straight away and the assistant knows which ones Amazon accepted; product changes are queued and stay pending until you publish them: they do not reach Amazon before that.
  • It will ask you before applying, just as before closing a task: both are announced to the assistant as operations that require confirmation, because applying really changes your campaigns and your listings, and closing cannot be undone.
  • Not every suggestion gets applied. Today, product suggestions and campaign ones -daily budget, state and name- are applied. The rest stay as notes for you to read: approving them changes nothing, and the assistant knows it.
  • Close the task at the end. Whatever is still pending when you close it stays frozen and can no longer be approved.

The tasks the assistant creates appear in your list with the origin API and assistants, and each decision records which credential made it.

Rewriting your listings with the Epinium AI ​

With the tasks:write permission the assistant can ask Epinium's AI to rewrite your Seller or Vendor listings: the title with its highlights, the bullets, the description and the search terms, 1 to 20 products at a time. It is the same AI as in the app, so it follows each product's keyword clusters, the prompts linked to them and your brand identity. It does so with the epinium_optimize_listings tool.

Three things worth knowing:

  • It leaves you suggestions, not changes. The result goes to a task, the onboarding one, in Processes → Tasks. The suggestions appear as they are generated, and the assistant shows them to you when they are ready.
  • Approve, apply and publish. You approve or reject each suggestion, in the conversation or in the app, and applying the task leaves the changes pending to publish: they do not reach Amazon until you publish them, from the app or from the chat.
  • It is free up to 20 products. The rewrite does not use your credits: Epinium pays for it, with a cap of 20 products per account in total. Every product launched counts, including a repeated one or one whose generation fails, so the assistant agrees the products and fields with you before launching it and does not repeat any unless you ask. Past the cap it will tell you there is no room left, without generating anything. Approving or rejecting the suggestions is billed as in any task.

Creating, editing and deleting keyword clusters ​

With the catalog:write permission the assistant can save what it finds as a keyword cluster, the same ones you manage in Segmentation → Clusters: ask it to group a few products with the keywords they should rank or advertise for, and it creates the cluster or extends one you already have. Nothing reaches Amazon: a cluster lives in Epinium, where your SEO prompts, the optimizer's keyword discovery and the cluster filter of the product charts pick it up.

Four things worth knowing:

  • It edits by deltas. It adds or removes keywords and products, or renames the cluster, without resending the whole list; a keyword it adds again is updated, not duplicated.
  • It respects your padlock. A cluster you protected from the app is refused, and the assistant knows it has to ask you to unprotect it rather than retry.
  • It can delete, and it will ask you first. Deleting a cluster cannot be undone, so the tool is announced as destructive and your AI client asks you to confirm. It is refused while the cluster is protected or a smart campaign uses it as its base.
  • It will not ask you to confirm creating or editing, precisely because no bid, budget or listing moves. The cluster is created under the name of the person who authorised the assistant, and it can link only Seller or Vendor catalog products of the account.

Writing the prompts your listings are written with ​

With the same catalog:write permission the assistant can draft your custom prompts, the ones in your prompts library, and tell each cluster which ones to use: ask it, for example, for a premium-tone prompt for the bullets and to link it to the cluster of your high-end range. The next time the content of that cluster's products is generated, Epinium's AI will follow those instructions.

Three things worth knowing:

  • A prompt changes nothing on its own. Saving it neither touches Amazon nor rewrites any listing: it applies when content is generated, and only for the products of the clusters it is linked to.
  • One prompt per field. Each prompt says which fields it applies to -title and highlights, description, bullets or search terms- or applies to all of them. If several prompts from a product's clusters apply to the same field, only one is used.
  • It creates and edits, but does not delete. To take a prompt off a cluster, unlinking it is enough; the prompt stays in your library. Prompts are created under the name of the person who authorised the assistant, and they count toward your plan's prompt quota.

Your brand identity and your competitors ​

With the catalog:write permission the assistant can create and edit your brand identity, the one in Settings → Brand: the name, the sector, your websites, your notes on tone and audience, and the catalog brands it covers. It is the context Epinium's AI uses when it writes your listings.

It also saves a profile per competitor, up to 50: its name, its website or store, the ASINs it competes with you on (up to 20), its price compared with yours (below, similar or above) and how dangerous it is (low, medium or high), always with the reason your data gives for it: shared searches, lost Buy Boxes, price or reviews. If there is no evidence yet, it leaves the threat level out.

Three things worth knowing:

  • Competitor profiles are not shown in the app yet. The assistant saves and reads them; Settings → Brand does not show them yet. The competitor websites you already had are kept separately.
  • When editing, it rewrites the whole list. To add a competitor it sends the existing ones plus the new one, so it reads the current ones first. The previous values are not kept anywhere, which is why your AI client asks you to confirm edits.
  • Nothing reaches Amazon. Saving the brand identity neither touches nor rewrites your listings, and the assistant cannot delete it.

Changing your Amazon listings, in the chat or with a spreadsheet ​

With the catalog:write permission the assistant saves changes to any attribute Amazon lets you edit in the product type of each listing, on Seller and on Vendor. You can ask for them in two ways, and both end the same:

  • In the chat, interactively: you tell it the change and it finds the products, asks you about anything unclear and prepares it. "Set the material to cotton on these 12 SKUs" or "shorten this product's title to 75 characters".
  • With a spreadsheet, in bulk: you hand it an Excel file or a CSV with one row per product. The material of 200 SKUs, the battery data missing from half your catalog or the bullet points of a whole range.

Nothing reaches Amazon without you. What it saves is left pending to publish, just like a change made by hand: you see it in the Properties tab of each product and in Publish changes in the catalog. It goes out to Amazon when you publish it from there or, if its key has catalog:publish, when the assistant shows you the list and you confirm it should publish it. Before it is saved it goes through the same validation as any other change.

How you hand it the changes ​

  • A spreadsheet (.xlsx, .xls or .csv). In Claude chats, on the web and desktop, and in ChatGPT, the assistant shows you a box to drop it into, and the file goes straight to Epinium without passing through the conversation. In Claude Code and in Cowork that box does not appear: the assistant reads the file and saves its rows itself, and with more than about 1,000 rows it will suggest doing it from a Claude chat, which is faster.
  • A message. "Set the material to cotton on these 12 SKUs" works too, even for every product in a country: the assistant looks them up if needed and turns your request into rows, without asking you for a file.

The three steps ​

  1. Prepare. It looks up the products by their SKU, ASIN or EAN in the marketplace you tell it, and checks which attributes they accept. With that it matches each column to an Amazon attribute, leaves out the ones that are only yours, such as cost or notes, and only asks you about the doubtful ones. If the spreadsheet has no country column, it asks you for the marketplace: it never guesses. If an ASIN or an EAN matches several SKUs, the change goes to all of them, and it tells you so. With EAN, listings whose EAN Epinium does not have come out as not found: use the SKU or the ASIN.
  2. Validate. It checks every row against Amazon's rules for its product type, without saving anything. It tells you what blocks and what is only a warning, with the values Amazon accepts, and if a value is wrong in many rows it offers to fix it in one go.
  3. Save. It validates everything again and saves the valid rows as pending changes. The ones that do not pass are not saved, and the assistant tells you which ones and why.

Which columns it understands ​

  • One column per attribute, with whatever header you give it ("Material", "Item weight"): the assistant matches it to the Amazon attribute.
  • Parts of an attribute. Some attributes hold several pieces of data, such as battery, with the cell composition and its weight. Each piece can go in its own column: a "Battery composition" column fills only battery.cell_composition.
  • Units. They come from the header, as in "Weight (kg)", or you tell the assistant.
  • Attributes with several entries, such as bullet points: one column per position, "Bullet point 1", "Bullet point 2" and so on.
  • Values go in as you write them. It understands the decimal comma and measurements such as 19 x 11 x 8, but it does not translate: text is saved in the language of the spreadsheet.
  • Price goes this way too, by audience: consumers and businesses. The rules and how to clear a value are in The price.

The price ​

  • What you can change. In Seller, for consumers (ALL) and for businesses (B2B): the price, the minimum, the maximum and the sale price with its start and end dates. Each price column is paired with an audience and a field. If the header does not say the audience, the assistant asks you.
  • The rules are the same for consumers and businesses. There is always a consumer price; the price stays between the minimum and the maximum; the minimum does not exceed the maximum; there is no minimum or maximum without a price; the sale stays between the minimum and the maximum and does not exceed the price (equal is fine); the sale start is before its end and the end is not before today; no amount is 0 or negative. A sale whose end has already passed counts as no sale.
  • Only what the row introduces blocks. A product that already broke a rule is not blocked for it: the row that adds the problem is blocked. The exception: if the row changes the price of a product whose saved price already breaks a rule (for example, a consumer price outside its minimum or maximum), it is blocked and that price has to be fixed in the app first.
  • Ending a sale early. Set its end date to today. The sale ends at 0:00 of the end date, so "Sale end" is the first day without the sale.
  • An empty cell never changes anything. To clear a value, the assistant asks you once per column ("there are 40 empty cells in "Minimum" on products that have a minimum: do I leave them or clear the minimum?") and only clears if you confirm.
  • What cannot be cleared. The consumer price cannot be cleared. Clearing the business price removes the whole business offer, with its minimum, its maximum and its sale.
  • An incomplete sale is blocked. A sale price with dates but no amount, or dates that are not real dates, are not saved. If the dates are ambiguous, the assistant asks you the order (day/month or month/day).
  • The currency is not converted. Amounts are saved in the marketplace currency; if the spreadsheet brings another one, the assistant asks you and never converts. An amount written as text with more than 2 decimals ("1.299 €") is rejected, so a thousands separator is not read as decimals; a numeric cell in the spreadsheet is rounded to 2 decimals, as you see it in the sheet.
  • The first price is set in the app. A product that does not yet have a consumer price in Epinium cannot receive one from a file, because the currency comes from that offer: set it from Prices.
  • It stays pending until you publish. Like the rest, the price change stays pending until Publish changes, and removing a sale price also reaches Amazon when you publish.

Data that go together ​

Some values force others to be filled in: if a product now includes batteries, Amazon also asks for their type, their composition and how many it has. Epinium detects those cases when it validates, and the assistant shows you each combination -the product type, the value that triggers it, what the spreadsheet brings and in how many rows- and asks you what to do with it:

  • If you confirm it in the chat, it is saved with the rest of the row.
  • If you do not confirm it, or the row is missing data, those changes go to a task in Processes → Tasks, named after the file, for a person to review and approve them. The rest of the row is saved anyway.

The assistant never treats a combination as confirmed unless you tell it so.

Several countries and Vendor ​

  • A spreadsheet with several countries. With a country column, each row goes to its own country. It recognises the code (ES, UK), the domain (amazon.es), the marketplace ID and the country name in its own language, in English or in Spanish. A row with a country it does not recognise is rejected: it never guesses. A country where you have no connection comes out with its rows as not found, without holding up the rest.
  • Each country is confirmed separately. Amazon's rules change from one marketplace to another, so confirming lithium batteries in Spain does not confirm them in France.
  • Each country, in its own language. Values are not translated: put "papel" in the row for Spain and "papier" in the row for France.
  • Vendor. With a vendor code column, each row goes to that vendor code. If a row without a vendor code matches listings from several vendor codes in a country, the assistant asks you whether to save the change in all of them; if you do not confirm, those rows are left out. In Vendor no price is edited from a file, neither the cost price nor the list price: those columns are left out and the rest of the row is saved.

Limits and credits ​

  • Up to 5,000 rows and 8 MB per file, all countries together, and up to 250 columns.
  • The file is kept for 7 days. To save its rows it needs at least one day left; otherwise, upload it again.
  • A file is saved once. It cannot be saved again with different settings: if something was not right, fix the spreadsheet and upload it again. If a product appears in several rows, the last one counts.
  • It costs 5 credits per row. At the start the file's rows are reserved and, if your balance falls short, nothing is saved. At the end only the rows that were saved or went to the task are charged, and the rest is refunded. Uploading, preparing and validating the spreadsheet cost nothing. The rates are in Settings → Credits.
  • It is for the Business and Master plans, like any write.

Listing new products on Amazon ​

With the catalog:write permission the assistant also puts up for sale products that are not in your Amazon account yet: an offer of yours on a product that already exists (on Seller), a new product or a variation family with its parent and its children, on Seller and on Vendor. It finds the product, checks it against Amazon's rules and saves it as a pending new listing, which goes out to Amazon when you publish it.

What it will ask you, how a family is published, what changes on Vendor and what it costs are in New product listings.

Review and publish from the chat ​

In Claude chats, web and desktop, and in ChatGPT, the assistant opens the Review and publish window inside the conversation. It is the same as Publish changes in the app:

  • Three tabs: Changes, New listings and Publications, with filters by connection and country and a search by SKU or ASIN.
  • You publish what you select, and in Publications you follow whether Amazon accepts it, including the new listings being imported into Epinium.
  • Publishing from it needs the catalog:publish permission (on the permissions screen, Publish to Amazon), besides catalog:write. Without it, the window shows you what is pending but does not let you publish. An authorization you already had does not get it by itself: authorize the assistant again.
  • In Claude Code and Cowork the window does not show. The assistant shows you what will go, by connection and country, and only publishes if you say yes.
  • Publishing cannot be undone. A mistake is fixed by publishing the previous value.

Writing a product's cost ​

With the own costs:write permission -not catalog:write, which is for clusters, prompts and listings- the assistant creates or updates a product's cost with epinium_set_product_cost: the same action you take on the product's own page or when you upload the cost CSV template, with the same semantics. For many products at once it uses the import.

Four things worth knowing:

  • The semantics are the CSV's. If you only ask it to change the unit cost, the shipping cost it already had stays untouched; a 0 sets it to zero for real, which is not the same as leaving it alone.
  • A past date recalculates the history. By default it writes as of today; if you ask it to apply the cost from an earlier date, the P&L of the periods already closed after it recalculates on its own, and it takes a little while to show up.
  • Only on the plans that include product cost. On Free and Guru this tool responds with a 403, same as every other write on those plans.
  • It does not change the cost price of a product with purchases. Its cost is calculated from the recorded purchases, so the assistant will tell you it cannot. It does change the shipping cost, as long as it sends it alone, without the cost price: with both at once, nothing is saved. To change its product cost, the assistant records or corrects a purchase (Purchases).

It will not ask you to confirm before writing -it is not a destructive operation, it deletes nothing and the previous value stays in the history-, but it will mention the cost and the date it applied.

Purchases ​

When a product has purchases, its product cost is the average cost of its stock: each purchase recalculates it with the units you had and the ones you buy. The assistant tells you which purchases a product has, with the average cost each one left, and which products it shares stock with; and it records, corrects and deletes purchases. It is the same as Record purchases on the product page, with the same calculation. To read them it needs catalog:read; to record, correct and delete them, costs:write. Both only on the plans that include product cost.

What is worth knowing:

  • Shared stock has a single cost. With Pan-European FBA, the same SKU in the EU countries is the same stock, so a purchase changes the cost of all of them (UK and Vendor are separate). The group is created with the first purchase; if those countries have different costs today, that purchase also needs what the stock you already had was worth, and without it it is not recorded.
  • The stock before the purchase. If you do not tell it how much stock you had, it takes the latest Amazon FBA stock from the previous 7 days, plus the purchases recorded since then. On FBM you have to give it.
  • The order number is not repeated. If the purchase has an order number, recording the same one again is rejected instead of duplicating it.
  • Correcting or deleting recalculates. It changes the average cost of that purchase and of the ones after it, and the P&L of the periods already closed, like a cost with a past date. Deleting cannot be undone; if it was the group's last purchase, the costs from before apply again.
  • It asks you to confirm before correcting or deleting. They are marked as operations that cannot be undone, so the assistant asks for your confirmation; recording a purchase does not.
  • With several accounts, reading needs one and writing needs all of them. If the stock is shared by accounts the key does not reach, the assistant can read their purchases but not change them.
  • Each write costs the same as writing a cost.

Importing the costs of many products ​

To load the cost of many products, the assistant uses epinium_import_product_costs, with the same costs:write permission. It is the same import as Import costs from a file in the app, with the same template and the same rules, and it recalculates the P&L only once, at the end. Product by product does not work for this: each cost with a past date recalculates the P&L of the whole account and, while it runs, the next ones are rejected.

What is worth knowing:

  • You pass it the file in a box to drop it in. In Claude chats, web and desktop, and in ChatGPT, it shows you the box, and the file goes straight to Epinium: .xlsx, .xls or .csv, up to 8 MB and 200,000 rows, so a whole catalog fits in one file, without splitting it. If an Excel file with many columns does not fit, remove the columns you do not use or save it as CSV. In Claude Code and Cowork it reads the file and passes it the rows itself, up to 5,000 per import.
  • It understands the columns like the app: the template's and common names like "sku", "cost" or "date". If it does not recognize one, it says which column each field comes from, without asking you for another file. Each row follows the rules of the app's file.
  • Before importing it shows you the check. Without saving or charging anything, it tells you how many rows would be applied, how many would change nothing because the product already had that cost on their date, how many would not be applied and why, how many would go without a date and how many credits it would cost. It only imports when you say yes, and if the check changes when you change the date, it shows it to you again first.
  • Dates matter. A row without a date applies from today and does not cover earlier sales. If your file has no dates, the assistant asks you from when the costs apply and uses that for those rows. And if a row has a date earlier than a cost the product already has, it tells you that row only counts until that cost, which stays.
  • Variation parents are rejected. They do not sell, so their cost would cover no sale: the cost goes on their variations.
  • It gives you the result when it finishes: how many rows were applied, how many changed nothing and how many were rejected, with the reason for each one. Until then it will not tell you the costs are loaded.
  • It stays in the app's history, in the History of the cost import of that account.
  • One account per import. If the P&L of that account is being recalculated because of an earlier cost change, it imports nothing: wait for it to finish and ask again.
  • Each applied row costs the same as writing the cost of one product. At the start the rows that can be applied are reserved and, if your balance is not enough, nothing is imported. At the end only the applied ones are charged and the rest is returned. Checking how it is going costs nothing. Only on the plans that include product cost.
  • Purchases too. It imports purchases from a purchases file, only on Seller accounts. Its check only reviews the format of each row and already rejects badly formatted ones: the SKU, its stock group and the stock are checked when importing, so a few more rows may be rejected then. Each applied purchase costs the same as recording it on its own.

Writing own expenses ​

With the same costs:write permission, the assistant manages your own expenses: the same ones you add in Costs and expenses → Own expenses. You can dictate them in the chat or hand it a spreadsheet with your expenses, and it creates them with epinium_create_expenses. With a spreadsheet, in Claude chats on the web and desktop and in ChatGPT it shows you a box to drop it in: the file goes straight to Epinium and is read with the rules of the app import, without the assistant copying its rows. If its columns are not named like the template, the assistant tells Epinium which one holds each field. In Claude Code and Cowork, which do not show the box, it reads the spreadsheet itself.

Five things worth knowing:

  • It first shows you what it is going to create. With several expenses, the assistant checks them beforehand without creating anything or spending credits, shows you the result and only creates them when you say yes. If any of them has an error none is created: it will tell you which one fails so you can fix it, or create only the valid ones if you ask it to.
  • It warns you about possible duplicates. If an expense looks like one you already have, or you put it twice in the spreadsheet, it asks you before creating it. And if an amount in the spreadsheet could be thousands or decimals, like 1.500, it asks you instead of guessing, just as the app import rejects it.
  • It asks you before changing an amount, ending or deleting an expense. A new amount applies from the month you choose, and the assistant will ask you from which one. Ending it deletes the amount changes it had scheduled after that. Deleting an expense removes it from the whole history; if you only want it to stop counting, ask it to end it.
  • The P&L subtracts them straight away. What you create, change or delete shows up in the next P&L query, including in months that have already passed.
  • Only on the plans that include product cost. On Free and Guru these tools respond with a 403.

Querying your Shopify and WooCommerce stores ​

With the platform:read permission the assistant reads, live, what is inside the Shopify and WooCommerce stores you have connected: products, variants, stock, prices and the rest of what each store exposes. It does so with two tools that work for both platforms:

  • epinium_platform_docs tells it which version of the store's API Epinium uses, where its reference lives and the pitfalls that make a response look right when it is not. It also shows it the fields your store has right now, including the ones your plugins or apps add, so it does not write them from memory.
  • epinium_platform_request runs the query in each platform's own language: GraphQL on Shopify and REST on WooCommerce.

With the platform:write permission, the same epinium_platform_request also creates, edits and deletes products and content in the store, directly and with no task. What it can change on each platform, and what gets recorded, is in the Shopify and WooCommerce guides. To copy your Amazon catalog to your store, step by step, see the use cases.

Three things worth knowing:

  • It is a different permission from connections:read. That one says what you have connected; this one lets it read what is inside the store. That is why it has its own area on the permissions screen, Data from your connected stores and platforms.
  • Personal data stays out. Names, emails, phones and addresses are refused before the query reaches the store. It is a deliberate safeguard, not a bug, and the assistant knows which field triggered it.
  • It is charged by the size of the response. Asking only for the fields you need is cheaper and faster.

What data it reaches ​

The same as the API, resource by resource. The full list of endpoints, with what each one returns, is in the API section.

Five things worth knowing when asking it for data:

  • Metrics need a date range. Tell it the period you care about; otherwise the assistant will ask.
  • You can ask for the day-by-day evolution, not just the total for the period. Useful for spotting trends rather than a single aggregate figure.
  • The range cap depends on the granularity. With the period total or by week, month, quarter or year, 366 days; day by day, 93. For targets and search terms, the heaviest resources, it drops to 93 and 31. Asking for more returns an error stating the exact limit.
  • An empty figure is not a zero. If the assistant says a metric is unavailable, it means Amazon does not report it for that ad format, not that it is zero.
  • Before telling you a metric does not exist, it looks for it. With epinium_find_data it finds the tool and the field of any metric by its name or its acronym. If it still says it does not have it, the MCP does not expose it.

Questions it answers well ​

  • Which campaigns have the worst ACOS this month?
  • Which searches am I spending on without any sales?
  • Which keywords are performing best and would be worth promoting to exact match?
  • Which advertised products are not getting a single impression?
  • Which of my products get traffic but do not convert?
  • What has the workflow changed on this campaign over the last month, and did it achieve anything?
  • Which campaigns does my workflow reach, and is there one I think it optimises that it never actually touches?
  • Which products in my Shopify or WooCommerce store are running out of stock?
  • How much did each of my clusters sell last month, and how much did I spend on advertising in each?
  • Which sales channel gives me the best real margin, after everything it costs me to make and sell the product?

Epinium Documentation