Not all Amazon Ads MCPs are built equal

There are two camps when it comes to using AI in retail media advertising.

1. The black box camp has handed their accounts to AI tools that decide and act on their own. They may save hours on manual work every week, but the trade-off they've accepted is that most of the time they can't say why a change was made, which rule/policy it followed, or what the tool will do next.

2. The sceptics (the other camp) looked at the same tools and said "no". All they see are changes being made without oversight, with no rules behind them and no record of the change, and because they're the ones answerable (and not the AI), they'd rather keep doing the work by hand than rely on something they can't verify or control.

Both camps have a fair point. The first recognises the obvious advantage of saving time on manual changes and optimizations that would normally take days or weeks of analysis and implementation, and the second is right that you can't be accountable for a system you can't see into.

3. The third camp wants an AI that works inside the policies they've already set, makes changes they can review, and shows its work. That's the standard they hold any assistant to once it connects to their ad account. They've already put real work into their strategy: a target ACoS set campaign by campaign, rulesets that encode how they bid, and dayparting for the hours that matter. They don't want AI to replace that setup. They want it to read the setup, question it, and help them improve it, at scale and with speed.

Their questions are about the policies as much as the campaigns. Why did ACoS climb on these campaigns, and which rule moved the bids? Why has this rule never fired? Is this ruleset switched on but linked to nothing? And when the assistant changes something, they want a record of what changed, who asked for it, and why.

A black box can't answer any of that because it never exposed its rules, and an assistant that only sees campaign numbers can't either.

What most ads MCPs get wrong

A typical ads MCP, including the ones the ad platforms themselves offer, hands over your raw data to AI lets it change bids, budgets, targets, or anything else... one at a time. That works well when you're managing a handful of campaigns.

Once you have dozens of campaigns across several accounts, it turns into a long list of individual edits that nobody on your team can realistically review, and the same problem tends to show up again the following week because the fundamentals never changed.

Behind every bid and budget in a well-run account, there's a policy that sets it: when to adjust bids against a target ACoS, how a dayparting schedule shifts budget through the day, or a threshold that caps how far a bid can move. + Hundreds of other policies that take years of judgement to make.

The success of a campaign is the output of those policies, and an AI that can only see the output doesn't understand why it's set up the way it is.

How the Adbrew MCP is built

Fortunately for Camp Three, Adbrew MCP connects your assistant to the policies, rulesets and thresholds that already run your account, so it works through your strategy instead of guessing/hallucinating.

The advantage is that a result can be traced to the rule behind it, and when you fix that rule, the fix applies to every campaign linked to it from its next run, rather than one campaign/account at a time.

Here's what that looks like:

• Reads your rules. It reads your rulesets and thresholds, which campaigns each one covers, how often it fires, and what it has changed.

• Suggests changes to them. Your assistant reviews your rules and suggests changes, and your team decides what to apply and edits the rule in Adbrew.

• Acts on campaigns. Your assistant updates keyword and target bids, target state, campaign budgets and state, and product ad state, and every change is recorded with who made it and why.

• Works across accounts. It reads performance and entity data for Sponsored Products, Sponsored Brands and Sponsored Display across every account you manage, making cross campaign/account analysis a breeze.

What it means for you

For the people running the accounts, the advantage of Adbrew MCP comes down to five things:

• Verifiable answers: When your assistant explains why performance changed, it points to the rule and the setting behind it, so you can open that rule and confirm it instead of just trusting it.

• One fix instead of hundreds: Correct a rule once, and the fix will apply to every campaign linked to it from the next run, whether that's ten campaigns or several hundred, even across accounts.

• Problems humans would otherwise miss: It can tell you that a rule has never fired, or that a ruleset is switched on but linked to nothing, and other misconfigurations that don't show up until the damage is done.

• Meaningful hours saved: An audit that takes days of research, exporting and cross-checking becomes a conversation worth a few minutes. And your team? They still decide which rule changes go live with every change made through the MCP, recorded with who made it and why.

• No rework: It works with the rulesets, thresholds, and targets you've already set up, so the strategy you've spent years refining is what the AI follows.

What that looks like in practice

Renee Cosmetics connected Adbrew MCP to Claude ...

"Before using it for our brand, I checked its answers against the Adbrew dashboard for spend, sales, and ACoS across a few accounts, and the numbers matched.

Since it pulls data directly from Adbrew rather than relying on a separate copy, I knew I could trust it. Now, I use it regularly to understand performance and make more informed decisions for our brand.”