MCP for marketers: what it is and how to use Claude connectors

TL;DR

• MCP (Model Context Protocol) is an open standard that lets AI apps like Claude and ChatGPT plug into your tools and data. Anthropic open-sourced it in November 2024.
• In Claude, an MCP server you add shows up as a connector. Custom connectors work on every plan, including Free, which gets one.
• For marketers the payoff is direct access: Claude reads your Search Console, your docs, and your CMS itself instead of relying on screenshots you paste in.
• Treat every connector like a new hire with keys to the building. Connect trusted servers only, and keep anything that writes or publishes on approval.

MCP lets Claude open your marketing tools itself, so you stop pasting exports into a chat window. Here is what it is, how connectors work, and where to start.

What is MCP, in plain English?

MCP stands for Model Context Protocol. The official documentation calls it an open-source standard for connecting AI applications to external systems, and compares it to a USB-C port: one shape of plug, lots of devices.

Before MCP, every AI app needed a custom integration for every tool it wanted to talk to. Anthropic's launch announcement described models as stuck behind information silos, with each new data source needing its own build. MCP swaps that mess for one shared protocol.

For a marketer, the change is easy to feel. Without it, Claude only knows what you paste in. With it, Claude can open Search Console, read the doc in your Drive, and check what's already live on your blog before it writes anything.

Where did MCP come from, and who supports it?

Anthropic open-sourced MCP on 25 November 2024. David Soria Parra and Justin Spahr-Summers created it, and it shipped with a specification, SDKs, and pre-built servers for tools like Google Drive, Slack, and GitHub.

It didn't stay a Claude thing for long. The MCP docs now list Claude, ChatGPT, Visual Studio Code, and Cursor among the apps that support it, so a server built once works across all of them.

Search demand shows how fast this moved. Google Keyword Planner puts "what is mcp" at around 49,500 searches a month worldwide, and "claude connectors" went from 1,900 searches in October 2025 to 22,200 in March 2026. That's a lot of people trying to work out the same thing you are.

How does MCP work, without the jargon?

Picture Claude as a sharp new hire on their first morning, fast and well read and completely clueless about your business. Every MCP server you connect is a login you hand them, one for Search Console, one for your Drive, one for your CMS.

The architecture docs split it into three roles. The host is the AI app you're using, like Claude. The client is the connection the host opens to each server. The server is the program that exposes a tool or a data source.

Servers offer three kinds of things. Tools are actions Claude can take, like pulling a report or saving a draft. Resources are data Claude can read, like a file or a database record. Prompts are reusable templates, the closest thing MCP has to an SOP.

Servers also come in two flavours. Local servers run on your own machine, while remote servers are hosted online and reached over the internet. If you're not technical, remote is the one you want, because setup is a URL and a sign-in.

What is the difference between an MCP server and a Claude connector?

A connector is Claude's name for an MCP server added to your account. Some come from Claude's connectors directory, and you add others yourself as custom connectors by pasting the server's URL.

According to Claude's help centre, custom connectors work on Free, Pro, Max, Team, and Enterprise plans, with Free accounts limited to one. On Pro or Max you go to Customize, then Connectors, add a custom connector, paste the URL, and sign in. On Team and Enterprise plans an Owner adds the connector first, then each person connects their own account.

One detail catches people out. Claude connects to remote servers from Anthropic's cloud, not from your laptop, so a server sitting behind a company VPN won't connect unless your IT team allowlists Anthropic's IP ranges.

What can a marketer do with MCP connectors?

Start with a real example. This post exists because of an MCP session. We connected the Contengi server to Claude and asked it to find topics our blog should cover and doesn't. It pulled 90 days of our Search Console queries, ran keyword volumes through Keyword Planner, and checked the results against roughly 170 posts we'd already written.

MCP came back near the top. We'd written about Claude Code, Claude skills, and the Claude API, and never once explained the protocol that ties them together. Doing that by hand means three exports, a spreadsheet, and a lost afternoon. It took one conversation.

The same pattern works for recurring marketing jobs. A weekly performance summary from Search Console and GA4, written in your words. Drafting straight into your CMS with your brand knowledge base loaded first. Refreshing posts that are slipping, using the live page and its search data instead of the model's memory. Turning call transcripts into first drafts that keep your phrasing.

Each of those runs on context. Claude writes better when it can see the real numbers and the real brand rules, and MCP is how it sees them.

Is MCP safe to use with your marketing data?

It's as safe as the servers you connect. Claude's help centre says plainly that custom connectors can reach services Anthropic hasn't verified, and once connected, Claude can read and change data in them based on your permissions.

The guidance boils down to a handful of habits. Only connect servers built by companies you trust. Read the permissions a server asks for when you sign in, and refuse anything that looks excessive. Watch for prompt injection, where a malicious server hides instructions meant to steer Claude. Save "Allow always" for tools you trust to run unsupervised.

One setting catches people out. In Claude's Research mode, Claude can call connector tools without asking each time, so Anthropic recommends switching off tools that write to other apps before you run it.

Our own rule is simple. Start every new connector on read-only jobs, and keep anything that publishes behind an approval step. You lose a few seconds and keep your live site safe.

Do you need to be technical to use MCP?

Not to use it. Adding a remote connector is about as hard as connecting Slack to Google Drive. Building a server is developer work, and you don't need to build one to benefit from one.

It gets harder when you stitch several servers into a workflow that runs the same way every week, with your brand rules applied every time. That's what content engineering is for, and it's where the hours go for small teams. Our guide to Claude Code for marketers covers the deeper end if you want to build it yourself.

Contengi does that wiring as one connector. You add it to your own Claude, and the knowledge base, Search Console and GA4 data, drafting, live-post editing, and playbooks arrive as more than 60 tools, running on your own Anthropic key.

How do you get started with MCP this week?

Keep it small. Pick one job you repeat every week, like the Monday traffic check or the monthly content report. Find a connector for the tool that holds the data, connect it, and run that job through Claude three weeks in a row before you add anything else.

Once it holds up, add a second connector and a write action. Less steps at the start means you spot problems while they're cheap to fix.

Frequently asked questions

What does MCP stand for?

Model Context Protocol. It's an open standard, created by Anthropic, for connecting AI apps to external tools and data.

Is MCP only for Claude?

No. The official docs list ChatGPT, Visual Studio Code, and Cursor alongside Claude as apps that support it.

Is MCP free to use?

The protocol is open source and free. The tools you connect may have their own pricing, and Claude's Free plan allows one custom connector.

What is the difference between MCP and an API?

An API is one app's own interface. MCP is a shared standard that lets an AI app discover what tools a server offers and call them, so one integration works across many AI apps.

Can Claude publish to my website through MCP?

Yes, if the server you connect has a publishing tool. Keep publishing behind an approval step so nothing goes live unreviewed.