The weekly client report is not the work.

The reporting tax

Client reporting is the clearest example of work that is necessary, repetitive and almost entirely mechanical. Log in, set the date range, export, repeat per platform, repeat per client, reconcile the figures that disagree, write the same four observations you wrote last week, attach the deck. None of it requires the judgement you are actually being paid for, and it eats the hours when that judgement is sharpest.

Muze does the retrieval and the arithmetic. compare_meta_performance and compare_google_performance both return this period against the previous one by default, so “what changed since last week” is one question rather than four exports. get_blended_performance puts total spend across Meta, Google and Amazon against Shopify revenue in a single call. You still write the recommendation, which is the part clients are paying for. You stop assembling the evidence by hand.

Audits on day one instead of week three

A new client account is a forensics job. Which campaigns are structurally broken, which search terms have been quietly burning budget for six months, which product feed is half disapproved, which ad sets have been live since a former employee set them up and nobody dared touch them. This is exactly where an agent earns its keep, because the work is broad, shallow and tedious rather than clever.

Muze reads all of it. list_amazon_search_terms surfaces pending harvest and negative recommendations with a row id on each. get_merchant_feed_health returns approved and disapproved product counts with the issues attached. list_campaigns_detailed pulls per-campaign delivery, budget and objective in bulk. inspect_google_campaign_ads goes down to the ad level. The audit becomes something you run in an hour on the day you get access, with findings in a form you can paste straight into a kickoff document.

The approval trail is the feature

Confirm-gated writes read like a safety feature. For an agency they are also a workflow feature. Every write returns a preview of the exact change before it executes, which gives you a concrete artefact to put in front of a client before their budget moves rather than an explanation afterwards. Campaigns are created paused, so you can build the entire structure, hand it over for sign-off, and enable it when the client says yes.

Tools that stop live spend are classified as destructive and require an explicit confirmation argument rather than a plausible instruction. On Amazon, list_amazon_autopilot_actions is a straight audit log of what the autopilot did on a profile and when, which is the answer to the question every client eventually asks about automation.

Never a percentage of spend

Tools and agencies that take a cut of ad spend get paid more when their clients spend more. We think that is a conflict of interest, so Muze charges for software and nothing else. There is a second reason this matters to you specifically: your cost does not scale with your clients’ budgets. Scaling an account does not scale your bill, and a client asking for a budget increase is not a line item you have to explain.

Muze also never holds, touches or intermediates ad budget. Spend runs through the client’s own ad accounts on the client’s own payment method, exactly as it does today.

What you would actually type

Account ids are resolved by the tools themselves. get_my_accounts, list_google_customers, list_amazon_profiles and list_shopify_stores each return what is inside the connection, so you name the client and ask.

• “Pull last week against the week before for this client on Meta and Google, and tell me what actually changed.”
• “Audit this Google account: which campaigns have no negatives, and which search terms are spending without converting?”
• “Check the Merchant Center feed. How many products are disapproved, and what are the issues?”
• “Draft the weekly summary: spend, blended ROAS against Shopify revenue, top three movers.”
• “Build a Search campaign for this landing page and leave it paused so the client can review it.”

What it cannot do without a human

• Every write previews first. The preview is the artefact you send the client. Nothing runs until someone confirms it.
• Campaigns are created paused. Build it, hand it over, enable it after sign-off. The paused status is forced by the API, not requested in a prompt.
• Destructive actions need an explicit confirmation. Pausing a live ad requires a confirm flag, so a confident-sounding instruction is not enough.
• Scoped OAuth. Access covers only the accounts authorized in the connection, and it can be revoked from the platform side.
• Autopilot keeps a log. list_amazon_autopilot_actions returns the audit trail of what changed on a profile and when.

What it costs

The free tier is 25 read-only tool calls per month, which is enough to run a genuine audit on one account before committing to anything. Write access requires a paid plan, and Muze is flat software pricing rather than a percentage of managed spend.

How you want client accounts separated across your team is worth an actual conversation rather than a pricing table. Start on the free tier, connect one account, and talk to us about the rest.

Frequently asked questions

Can Muze handle more than one client account?

The connection covers every ad account you authorize. get_my_accounts, list_google_customers, list_amazon_profiles and list_shopify_stores each return the accounts inside a connection, so you can name a client and ask about their account in the same session. If you need strict separation per client, connect them separately and talk to us about how to set that up.

Will it change a client’s account without asking?

No. Every write returns a preview and waits for explicit confirmation. Campaigns are created paused. Tools that stop live spend require a confirmation flag rather than an inferred instruction.

Can I show a client what it is about to do?

Yes. That is what the preview is for. The change is described before it executes, so approval happens on the specific action rather than on a general permission.

Does Muze take a cut of the ad spend we manage?

Never. Muze charges for software. Budget runs through the client’s own ad accounts, and Muze never holds or intermediates money.

Does this replace our reporting platform?

It replaces the retrieval and reconciliation step, not a white-labelled client dashboard. Muze has no branded report builder or scheduled PDF delivery. If clients expect those, keep your reporting tool and use Muze to get to the numbers and the actions faster.

Which platforms does it cover?

Meta Ads, Google Ads, Amazon Ads and Shopify through one connection. The wider Muze platform also covers TikTok Ads and Bing Ads, but those are not reachable through the MCP.

What does it cost?

The free tier includes 25 read-only tool calls per month. Write actions require a paid plan. There is no percentage of ad spend on any plan.

Keep reading

Other segments