If you've heard the term “MCP” and nodded along without being completely sure what it meant, you're in good company. This article explains what an MCP is in general, what the Perpetua MCP does specifically, and the kinds of questions it's genuinely good at answering, so you can decide whether it's worth your time before you set anything up.
Quick links
What is an MCP?
MCP stands for Model Context Protocol. It's an open standard, not a Perpetua product or a Claude feature, that gives AI assistants a safe, structured way to reach data and tools that live outside them.
The problem it solves is simple enough. An AI assistant on its own knows a lot about the world in general and nothing at all about your business. It has never seen your spend, your ACoS, or last Tuesday. Historically, the only way to close that gap was to paste your data in yourself, which is tedious, error-prone, and doesn't scale past a screenshot.
An MCP connector closes the gap properly. Think of it as a plug: on one side is an AI assistant, on the other is a system that holds real data, and the protocol is the shape of the plug that lets any assistant talk to any system without a custom integration for every pairing.
Two things follow from that, and both are worth holding onto:
Connectors are per-system. A Perpetua MCP gives an assistant access to Perpetua data. It says nothing about your email, your files, or anything else.
Permissions come from the system, not the assistant. The connector can only surface what your access to that system is allowed to see.
Perpetua’s own MCP
The Perpetua MCP is Perpetua's own connector. Once it's switched on, you can ask the AI assistant questions about your advertising data in plain English, and the answers come back from your actual Perpetua data rather than from the assistant's general knowledge.
A few specifics that shape how it behaves:
It reads your data. It doesn't change anything. Our first MCP release can't pause a campaign, move a budget, or adjust a target. It's a question-answering tool, not a control panel.
Your Perpetua login decides what it sees. It's exactly what you already see in the Perpetua app: the same geo companies, the same permissions. A geo company, in Perpetua terms, is one advertising account on one marketplace, so Amazon US and Amazon CA count as two.
It works with any assistant that supports MCP connectors. The current beta runs on Claude, and our setup guide covers Claude, but the standard is open.
Answers take about 30 seconds. It's querying live data, not recalling something. That pause is the tool working.
Why this is different from asking an AI assistant about Amazon ads?
Ask an AI assistant a question about Amazon advertising without a connector, and you get general knowledge. That's useful for questions like “what does ACoS mean” or “how does Sponsored Brands bidding generally work,” but the assistant doesn't have the context of your business, so treat any figures it gives you as a guess, not a fact.
With the Perpetua MCP switched on, that gap closes. Ask the same assistant a question about your performance, such as “what's my ACoS trend,” and the figures come straight from your data in Perpetua rather than a guess. The general knowledge is still there to help explain what it found, but the numbers themselves are real and there’s actual context behind them.
What you can do with the Perpetua MCP?
Click through each of these phrases below and read through each scenario:
#1: Remove the friction of getting Perpetua data into AI
#1: Remove the friction of getting Perpetua data into AI
Plan seasonality based on your own calendar cadence, not around your reporting restrictions.
A brand runs its business on a fiscal calendar & brand grouping that doesn’t match any report template. Every year, peak season planning happens against whatever shape the reporting offers — calendar quarters, platform-default groupings — and the team mentally adjusts. With the MCP they ask for the comparison the way their business is actually structured: their period against the same period last year, their brand groupings, their definition of the season.
What it matters: The friction that matters isn't cost, it's selection. When you need a non-standard view of your business, you’ll require someone to build it. If it cannot be built, you end up planning your business against the shape its tooling offers rather than the shape it operates in. That's a strategic distortion, and planning around your peak season is where it's most expensive, because it's the year's largest budget commitment.
Available today with Perpetua’s MCP.
#2: Cover every brand and marketplace in one question
#2: Cover every brand and marketplace in one question
Monday triage across the portfolio.
An agency lead asks one ranked question across every Perpetua account they can access — spend change and sales change between two complete windows, with the underlying numbers — and uses it to decide where the team spends the week.
Why it matters: The lead is choosing triage over portfolio reporting deliberately. Reporting is an output; triage is a decision, and it's the one an agency lead is typically held accountable to. Today attention gets allocated by whoever escalated loudest or whichever client had a call that week, because checking all of them properly costs more than it's worth. This makes evidence cheaper and quicker to obtain. That changes what the team works on, which is a different order of value from making a report faster.
Available today with Perpetua’s MCP.
#3: Give AI the full context behind your advertising
#3: Give AI the full context behind your advertising
Deciding whether an ACoS drift needs intervention.
ACoS has drifted on a specific Perpetua goal. You ask your AI assistant with Perpetua’s MCP connected to walk you through what goal’s setup — goal type, ad unit, targets — to locate exactly where the movement sits, then asks the question that actually matters: is this the goal’s configuration doing what we set it up to do, or is this something new?
Why this matters: The costly error in retail media isn't missing a drift, it's intervening on one that was behaving correctly. A team that manually retargets a goal that was tracking to its intended strategy makes performance worse and then can't tell why. Separating "the market moved" from "our own configuration is working as designed" is the judgement call this value aims to deliver, and it's the one that most often gets made on instinct currently.
Partially available today with Perpetua’s MCP. The configuration layer — goals, targets, segments, the retail data underneath — that tells you whether the drift was intended, is not reachable yet on this first release. So today this use case locates the change; the full version will let you conclude something about it.
#4: Bring proprietary retail media intelligence into AI.
#4: Bring proprietary retail media intelligence into AI.
Defending the ad budget in a margin conversation.
Your finance team proposes cutting retail media spend by 20%. With the help of our MCP, your team considers what that cut would actually cost — using incrementality and new-to-brand data rather than just an ACoS number — and brings an answer to the meeting rather than a position.
Why this matters: Ad budget is the easiest line to cut and the hardest to defend, because ACoS is a ratio and a CFO is asking about absolute contribution. "Our ACoS is good" has never won that argument. Incrementality is the only measurement that answers the question being asked and Return on Consumer data is the only one that speaks in the language the rest of the business uses. This is where our proprietary measurements stop being interesting and become decisive for your business — and it's a conversation that happens on a schedule, every planning cycle, whether or not the team is ready for it.
Not available in this release. ROC, NTB, incrementality and AMC analytics aren't available through our MCP yet.
5: Bring Perpetua's expertise into broader business analysis
5: Bring Perpetua's expertise into broader business analysis
The retail media section of a board or QBR deck.
Someone preparing for a quarterly review brings their own context into the conversation — the inventory position, the margin picture, the growth forecast the business is working to — pulls the advertising performance from Perpetua using our MCP, and has the assistant reason across both, then format it as a summary and table for presentation.
Why this matters: The strategic shift is that retail media stops being an advertising report presented to the business and becomes part of the business narrative. That's a positioning change inside your own organisation, and it's what determines whether retail media is treated as a cost line or a growth lever.
Available today with Perpetua’s MCP, with a caveat. The financial and operational context doesn't have to come from Perpetua, you’ll supply your own margin and inventory picture, MCP supplies the advertising performance, your AI assistant reasons across both.
A word on all of this: when the assistant formats your data into a table, chart, or summary, that's the assistant doing the writing with what the Perpetua MCP gave it. Read it the way you'd read your own export, especially before it reaches a client.
What the Perpetua MCP doesn’t do.
We want to be upfront so you know where the current edges are. Currently, our MCP doessn't allow your AI assist to write or change anything in Perpetua. No pausing, no budget moves, no target changes. It is Read-only, by design, for this release.
There's data it can't reach yet:
Your Perpetua configuration, including goals, segments, targets, budgets, and branded, competitor, or category splits
Retail data, including inventory, pricing, buy box, and returns
AMC analytics, market share, incrementality, and Return on Consumer
Product tags and product families
Budget plans
For any of those, the Perpetua app is still the place.
It can't tell you why Perpetua's engine changed a bid. It can show you that a bid changed and what happened around it, but the engine's reasoning isn't in what it can see. Ask “why” and you may still get a confident-sounding answer, but that answer isn't coming from Perpetua. Your CSM is the right person for those questions.
It's still early-stage. A wrong answer can still look right: formatted, tabled, and plausible. Check figures against the Perpetua app before they reach a client or team member, and if something doesn't match, tell your CSM or our Support team and include the question you asked.
Who it’s for and how to get access?
The Perpetua MCP suits anyone who asks questions of their advertising data regularly: in-house teams checking performance between reports, and agency or multi-brand teams who want to look across everything they manage in one go. You don't need to be technical. If you can write the question, you can use it.
Right now it's in closed beta, which means access is switched on for named people rather than everyone. If you'd like to try it, ask your CSM, who can tell you whether you're set up and what the beta involves.
Once you have access, the setup guide walks you through connecting it, and takes about five minutes.
Related articles