ITBook your free audit

What Is MCP in Marketing? Revolution or Hallucination

AI & UpdatesBy Angelo Doni5 min read

Typographic cover of the article on MCP in marketing: real or hype?

Key takeaways

MCP (Model Context Protocol) is an open-source standard, released by Anthropic in late 2024, that lets an AI model connect to external tools and data in a standardized way. In marketing it promises one assistant that can read your ad accounts, design tools and email platforms from a single conversation. It will change how performance marketing works, but giving AI access to data is not the same as giving it the context of your business.

  • MCP (Model Context Protocol) is an open-source standard for connecting AI models to external tools: think of it as USB-C for LLMs.
  • It tackles the N x M problem: instead of a custom integration for every model and every tool, one shared protocol.
  • For marketers, the promise is one AI that queries Meta Ads, opens Canva and prepares an email, without switching tabs or exporting CSVs.
  • The real risk is confusing access to data with understanding the business: metrics like CPA, CTR, ROAS and CPC do not carry your positioning, margins or logistics.
  • MCP will mean less manual technical work and more strategy and creative work. It will not replace strategy, positioning or the human spark behind a good ad.

MCP in marketing means connecting an AI model directly to the tools you already use: ad platforms, design tools, CRMs, email platforms, dashboards. MCP stands for Model Context Protocol, an open standard that lets a large language model (LLM) talk to external systems without a custom integration for each one.

Is it the next big shift in advertising or a new bubble? Our answer: both, depending on how you use it. It will change a lot of the daily work. It will not replace the thinking.

What is MCP, in simple terms?

Over the last few months, one acronym has kept coming up in conversations between developers, advertisers and AI companies: MCP. A technical, almost boring name. It could still become one of the most important standards of the next technology cycle.

Anthropic released MCP in late 2024 as an open-source protocol for connecting AI models to external systems. Until then, every AI needed its own custom connection to every platform. The idea now is a common standard. The official MCP documentation uses the same comparison we like: a USB-C port, but for AI.

The stated goal is to solve the “N x M” problem: N AI models that each need to connect to M different tools. With MCP, in theory, a single model can talk to:

  • Meta Ads
  • Canva
  • Slack
  • Figma
  • CRMs
  • email platforms
  • databases
  • dashboards
  • internal software

Anthropic has been showing deeper and deeper integrations with tools like Canva, Slack and other productivity ecosystems. It is easy to see why performance marketers are excited, and why searches like “Meta Ads MCP” and “Google Ads MCP” are taking off.

The marketer’s dream: one AI that sees everything

Picture this. You open Claude or ChatGPT and ask:

“Which Meta campaign had the best ROAS in the last 30 days?”

Then:

“Create 3 creative variants for the segment with the highest CTR.”

Then:

“Open Canva, generate a new graphic and prepare the email for our VIP customers.”

All in the same place. No switching tabs. No exporting CSVs. No copy and paste.

This is exactly where MCP gets interesting. We are no longer talking about AI that generates text. We are talking about AI that orchestrates tools, data and workflows.

Where it breaks: data without context

This is probably the most important part of the article. More and more often we hear this:

“I showed the data to the AI and it told me that…”

Good. Great, actually. The drive behind that sentence is healthy: business owners and marketers want to understand more, read their own numbers and depend less on dashboards nobody can make sense of.

But there is a problem: the AI sees the data. It does not always see the context. And in marketing, context is everything.

We build a business strategy together. We define positioning, margins, creative saturation, customer care capacity, logistics limits. And then we reduce all of it to:

  • CPA
  • CTR
  • ROAS
  • CPC

Yes, but no. (If ROAS is the number you lean on most, we explain why it is not enough in MER vs ROAS.)

That is the real risk of the new MCP ecosystems: confusing the availability of data with a real understanding of the business.

Why data will not save weak creative

Over the last few years advertising has changed radically. The competitive edge used to be technical:

  • pixels
  • granular audience segments
  • hacks
  • complex setups

Today most of the distribution is automated. AI is taking over the operational side. So what is left for us?

The most important role: the advertiser’s.

Creative today is not just “a nice video”. Creative is targeting. When we work on hooks, messaging angles, pacing, editing and differentiation, we are telling the algorithm who our ideal customer is. Creative has become a signal for the machine. It is the core of how we run Meta Ads for e-commerce brands.

That is why MCPs are, all at once:

  • extremely powerful
  • dangerous
  • overrated
  • inevitable

Data-first or ideas-first?

Spending time with business owners from very different industries, one thing stands out. Many successful people have vision, have energy, bring others along, and start from ideas.

Ermethe is a performance agency and we work with data every day. But “data-driven” is turning into a religion. As if data could replace strategy, intuition, sensitivity, experience and creativity.

It does not work that way. Data is there to validate intuition, not to replace it. That order is built into our method.

When we flip it, something often works in the short term. In the long term we hit a wall. And that is where MCP risks becoming a bubble: when we assume that connecting everything automatically means understanding everything.

So will MCP really change performance marketing?

Yes. A lot. It will bring:

  • centralized operations
  • faster access to data
  • more advanced automation
  • better integrated AI agents
  • much faster workflows

But no, it will not replace:

  • strategy
  • positioning
  • understanding the customer
  • the ability to communicate

The future will probably look like this: less manual technical work, more strategic and creative work. And honestly? We love that. We are back to the roots of communication, with the tools of the future.

AI, big data and MCP will be incredible allies for scaling. But the spark that decides whether someone clicks or keeps scrolling is still deeply human.

The bottom line

MCP could become a huge standard over the next few years. Anthropic opened it up to the market and the ecosystem is growing fast.

But the more tools we have, the more context matters. Having all the data in the world without a strategy is still a great way to make wrong decisions very quickly.

In the performance marketing of the future, the winners will probably be the ones who do something very simple: use AI without ever stopping to think like a human being.

If you want a second pair of eyes on how your data and your strategy fit together, book a free 20-minute audit with us.

FAQ

What is MCP in marketing?

MCP (Model Context Protocol) is an open-source protocol that connects AI models like Claude or ChatGPT to external tools. In marketing it means an AI assistant can work directly with ad platforms, design tools, CRMs, email platforms, databases and dashboards, instead of you copying data in and out by hand.

Who created the Model Context Protocol?

Anthropic released MCP as an open-source standard in late 2024. Since then the ecosystem of tools that support it has been growing quickly.

Can MCP replace a marketing team or agency?

No. MCP speeds up access to data, automation and workflows, but it does not replace strategy, positioning, understanding of the customer or the ability to communicate. Those are still human jobs.

What is the biggest risk of using AI with MCP for advertising decisions?

Mistaking data for context. The AI sees the numbers, but not always the strategy behind them: positioning, margins, creative fatigue, customer care capacity and logistics limits. Decisions based on metrics alone can be wrong, and with MCP they can be wrong faster.

Sources

#MCP#AI#performance marketing#advertising

Want to apply this to your e-commerce?

A free consultation to see what would work in your store.

Book a free consultation