The questions worth asking do not fit inside one platform.

The question that does not fit in one dashboard

Single-platform tools are a trap for anyone doing real volume, because the decisions that matter are comparative. Is Meta stealing credit from Google. What is blended return against actual Shopify orders. Where is the marginal dollar best spent this week. Is Amazon growing total demand or cannibalising your own store.

None of those can be answered from inside one channel, and none of them can be answered by an assistant that has to be told to switch tools, re-authenticate, and forget the context it had two minutes ago. The reason Muze did not stop at one platform is that the second platform is where the tool stops being a viewer and starts being useful.

One connection, four platforms

The session starts with get_my_accounts, which returns your Meta ad accounts and Google customer IDs together. Its description tells the assistant to call it first and never ask you for IDs, which is a small thing that changes the feel of the whole product: the conversation starts with your question rather than with you copying account numbers out of a browser tab.

The other platforms follow the same pattern. list_google_customers returns every Google Ads customer the OAuth grant can reach, flagging which are manager accounts so the assistant picks a real client account. list_amazon_profiles resolves profile IDs along with marketplace, currency and account type, and it too is documented as the call to make first. list_shopify_stores resolves the store domain.

One OAuth grant covers all of it. No developer account, no API tokens to mint, no platform passwords handed over.

What you can actually compare

Comparison only works if the answers have the same shape, which is why compare_meta_performance and compare_google_performance are deliberate mirrors of each other. Both default to month-to-date against the same span last month and both accept a custom window, so an assistant can put Meta and Google side by side without inventing arithmetic in between.

Around them: get_meta_last30days gives a fixed rolling window when you need a stable denominator, get_meta_performance and get_google_performance give the headline metrics, and get_amazon_performance returns a daily series with spend, sales, orders, ACOS, ROAS, CTR and CVR per day. list_account_ads and list_google_ads_data give the assistant each platform's full working set when the question needs entity-level detail rather than totals.

Then get_blended_performance closes the loop by totalling all three against Shopify revenue in a single cheap read.

The limits, stated up front

Putting four platforms in one session does not merge their attribution models, and we are not going to imply that it does. Each platform still counts conversions its own way, over its own attribution window, and those windows overlap. That is the reason blended reporting exists and the reason we do not publish a unified conversion number.

Two more limits, both surfaced in the tool responses rather than hidden. Amazon figures on this connection are ad-attributed only, because Muze holds the Amazon Ads campaign management scope and not SP-API, so there is no marketplace revenue figure. And cross-platform totals are summed without currency conversion, so they are only meaningful when your accounts share a currency.

None of that makes the comparison useless. It makes it a directional tool for deciding where to look and where to move budget, rather than a courtroom exhibit.

Why one connection rather than four servers

The fair version of this argument includes the case against us. If Meta is your only channel, use Meta's own official MCP server. It is free, it is first-party, and it takes about five minutes. Google's official server is read-only, which is genuinely the safest thing to point an assistant at if you never want it to act. Both are reasonable choices and we are not going to pretend otherwise.

The case for Muze starts at the second platform. One connection means one authentication, one set of conventions, and one session where the assistant holds Meta, Google, Amazon and Shopify context at the same time. It also means writes with a consistent safety model across all four, rather than four different answers to the question of what happens when the model gets it wrong. We will still be here when you add Google.

What to type

These are prompts, not commands. Any MCP client will do: ChatGPT, Claude, Cursor or anything else that speaks the protocol.

• “Compare Meta and Google for this month against last month and tell me which is trending worse.”
• “Where should the next $1,000 a day go across Meta, Google and Amazon?”
• “Show me spend on all three ad platforms next to actual Shopify revenue for the last 30 days.”
• “Is my Amazon ACOS better or worse than what Google is returning right now?”
• “List every ad account you can see and give me a one-line health read on each.”

The tools behind it

Cross-platform analysis is entirely read-only. Nothing on this page can change an account, which is why it is the right place to start with a new connection.

• get_my_accounts (read): Meta accounts and Google customer IDs in one call. Always the first call.
• compare_meta_performance (read): Meta period over period, month-to-date by default.
• compare_google_performance (read): The deliberate mirror, so the two answers line up.
• get_amazon_performance (read): Daily Amazon series: spend, sales, orders, ACOS, ROAS, CTR, CVR.
• get_blended_performance (read): All three platforms against Shopify revenue in one cheap read.
• list_amazon_profiles (read): Profile IDs with marketplace and currency, resolved for you.
• list_shopify_stores (read): Connected stores, so the assistant never asks for your domain.

Platforms covered. Meta Ads, Google Ads, Amazon Ads and Shopify. Four, and only four.

Nothing spends without you

• Every tool in this workflow is a read. Nothing here changes an ad account.
• One OAuth grant, scoped to the accounts you authorize, revocable at any time.
• The free tier's 25 monthly read-only calls are enough to run a genuine cross-platform review.
• Where the data has limits, currency, attribution, Amazon scope, the tools report the caveat instead of smoothing it over.

Frequently asked questions

Can one AI assistant manage Meta, Google and Amazon ads together?

Yes. Muze exposes all three plus Shopify through a single MCP connection, so ChatGPT, Claude or any MCP client holds every platform in the same session. 149 tools in total: 8 core, 37 Meta, 41 Google, 48 Amazon and 15 Shopify.

Which platforms does the Muze MCP support?

Four: Meta Ads, Google Ads, Amazon Ads and Shopify. It does not support TikTok, Bing, LinkedIn or X. The wider Muze platform covers more channels, but the MCP is these four.

Do I need separate logins for each platform?

No. One OAuth connection covers all four, with no developer account, no API tokens to mint and no platform passwords. Your assistant resolves the account IDs itself rather than asking you to paste them.

Does it merge conversions across platforms?

No, and nobody honestly can. Each platform counts conversions its own way over its own attribution window, and those windows overlap. Muze gives you matching per-platform comparisons for diagnosis and a blended spend-against-revenue figure for the decision.

Should I use this if I only advertise on Meta?

Probably not at first. If Meta is your only channel, Meta's own official MCP server is free, first-party and takes five minutes. The case for Muze starts when you add a second platform and the interesting questions become comparative ones.

Is cross-platform analysis safe to run on a live account?

It is read-only. Every tool involved in the comparison reads data and none of them change anything, so you can run a full review on the free tier without write access anywhere near your accounts.

Read next

Other features

Or see every Muze feature.