You keep seeing "MCP" in AI threads and newsletters. Everyone talks about it like you already know what it means. Here is the short version, then we will set it up.
MCP in 30 seconds
MCP (Model Context Protocol) is an open standard that lets AI assistants connect to external tools and data sources. Before MCP, you copied data into ChatGPT, Gemini, or Claude and asked questions about it. With MCP, your AI assistant pulls the data itself, live, from the actual platform APIs. Our AI marketing and MCP glossary defines all the terms you will encounter along the way.
Think of it like this: without MCP, you are the messenger between your tools and your AI. With MCP, they talk directly.
How MCP Works
One question flows through four layers -- you only interact with the first
1
You
Ask a question in plain language
2
Claude
Decides which tools to call and synthesises results
3
MCP Server (Ooty)
Connects to the platform API with your credentials
4
Platform API
Google, Meta, YouTube, Amazon -- real data, live
Your credentials stay safe. OAuth tokens live on Ooty's encrypted servers. Claude never sees your passwords or API keys -- it only receives the data that comes back.
Why marketers should care
You already spend hours each week switching between dashboards. Google Analytics, Search Console, YouTube Studio, Meta Ads, Amazon Seller Central. Each with its own reports, exports, and quirks.
MCP collapses that. Instead of pulling a CSV from Google Ads, pasting it into a chat, and asking "what happened?", you just ask "what happened?" and the AI goes and gets the data.
One question, live data, real analysis. No tab switching. No exports.
The Workflow Shift
What changes when your AI can talk directly to your marketing platforms
Before MCP
1Log into Google Analytics
2Log into Search Console
3Log into Google Ads
4Log into YouTube Studio
5Export CSVs from each platform
6Paste into spreadsheet
7Cross-reference manually
8Write up your analysis
Total: 45-90 minutes
With MCP
1Open Claude
2Ask your question
3Get analysis with live data
Total: 2-5 minutes
What you need
An AI assistant that supports MCP. ChatGPT (Pro, Team, Enterprise, or Edu with Developer Mode on), Claude on any paid plan (web, desktop, or mobile), Gemini (paid), Cursor, Windsurf, VS Code, Cline, Continue, Goose, the Gemini CLI, or Kiro. Two things to know about plan gates: ChatGPT Free and Plus cannot add MCP connectors today, and Claude Free does not ship with the Connectors panel either. Goose, Gemini CLI, Cursor, Windsurf, VS Code, Cline, and Continue are all free to use, so there is always a way in. The steps below show ChatGPT and Claude side by side. The concept is identical everywhere.
An Ooty account. Sign up at ooty.io and pick whichever product matches your work.
Ooty connects your marketing platforms to your AI assistant. Ask ChatGPT, Claude, or Gemini a question and get an answer pulled from Google Analytics, Search Console, Google Ads, Meta, YouTube, Amazon, HubSpot, and more. No exports, no dashboards, no tab switc
Most marketing teams use AI the same way: copy data from one dashboard, paste it into ChatGPT, Gemini, or Claude, ask a question, and hope the answer is useful. The AI does its best with a screenshot or a CSV export, but it is working with a snapshot, not your
AI marketing has its own vocabulary, and MCP adds another layer. This glossary defines the terms you'll encounter when working with AI marketing tools, reading about them, or setting up MCP connections.
Terms are alphabetical. Where something is commonly confu
A license key. You get one after signing up. It authenticates your requests.
That is it. No developer tools. No terminal. Nothing to install beyond the AI client you probably already have.
Choose Your Starting Point
Each product connects Claude to a different marketing platform. Start with the one that solves your most immediate problem.
After choosing a product, you will do a quick OAuth flow (the "Sign in with Google" button you have clicked a hundred times). Ooty gets read access to your data. It cannot make changes on your behalf.
3. Add Ooty to your AI assistant
For most people, this is a URL paste. The exact path depends on the client.
In ChatGPT (Pro, Team, Enterprise, Edu): Open Settings, enable Developer Mode, then go to Connectors, Add custom connector. Paste https://ooty.io/api/mcp/seo as the server URL and your license key as the Bearer token. Save.
In Claude (web, desktop, or mobile): Open Settings, go to Connectors, Add custom connector. Paste https://ooty.io/api/mcp/seo and your license key. Claude Desktop uses the same Connectors UI as Claude web now. The old claude_desktop_config.json file does not accept remote MCP servers.
In Gemini: Settings, Connectors, Add custom connector. Same URL, same license key.
In Cursor, Windsurf, VS Code, Cline, Continue, or the Gemini CLI: Your client has an MCP config file. Add this snippet.
Swap seo for whichever product you chose (ads, social, video, analytics, commerce, crm). Swap in your real license key. Save. Restart the client if it does not hot-reload.
Your Ooty dashboard has a Setup tab with the exact instructions for your chosen client, ready to copy, so you do not have to hand-type any of this.
Setup Timeline
From zero to your first live data query in 8 minutes
8
minutes total
One-time setup. No developer required.
1Sign up at ooty.io2 min
2Choose your product1 min
3Connect via OAuth2 min
4Copy the config snippet1 min
5Paste into Claude settings2 min
6Ask your first question0 min
4. Verify it works
Open a new conversation and type:
What tools do you have available?
You should see Ooty's tools listed. If they appear, you are connected.
Your first real query
Now ask something that uses live data.
If you connected SEO:
What are my top 10 search queries by clicks in the last 30 days?
If you connected Video:
How did my YouTube channel perform last month compared to the month before?
If you connected Ads:
Give me an overview of my Google Ads account for the last 30 days. Spend, conversions, ROAS.
If you connected Analytics:
Summarise my website traffic last month. Top pages, traffic sources, any notable changes.
That is live data from your real accounts. Not a demo.
A workflow worth saving
Once you are comfortable, try a repeatable prompt. Here is a weekly health check you can reuse every Monday:
Weekly check-in:
1. Top 5 search queries by clicks this week
2. Did impressions go up or down vs last week?
3. Any queries where my position shifted more than 5 spots?
4. Core Web Vitals status for my top pages
Takes 30 seconds to type. Returns a grounded answer in about a minute. No dashboards opened.
Common questions
Is my data safe? Ooty sits between your AI assistant and the platform APIs. Your OAuth tokens stay on Ooty's encrypted servers and never reach the AI. The AI sees the data Ooty returns, not your credentials.
Can I use multiple products at once? Yes. Add multiple entries to your config. They all work in the same conversation. "Compare my organic search traffic to my paid ad performance" uses both SEO and Ads data in one answer.
What about clients? Connect each client account separately during OAuth. In conversation, specify which account: "Using [client name]'s Search Console data..."
Does this work with ChatGPT and Gemini, not just Claude? Yes. MCP is an open standard. Every major client supports it: ChatGPT (Developer Mode on Pro, Team, Enterprise, or Edu), Claude (any paid plan, across web, desktop, and mobile), Gemini (paid), Cursor, Windsurf, VS Code, Cline, Continue, Goose, the Gemini CLI, and Kiro. ChatGPT Free, ChatGPT Plus, and Claude Free cannot add MCP connectors today. The free developer clients (Cursor, VS Code, Cline, Goose, Gemini CLI) always work.
What does it cost? Ooty products are licensed per product, with an all-access bundle available. No per-query fees. No usage limits.
What to try next
Once the basics feel natural:
Connect a second product and ask cross-platform questions. "How does my organic traffic trend compare to paid?" needs both SEO and Analytics data.
Build reporting prompts. "Write a 200-word executive summary of last month's SEO performance" gives you something you can drop straight into a client report.
Ask for anomalies. "Anything unusual in my data this week compared to the four-week average?" catches things you miss when looking at the same numbers daily.
The setup is a one-time task. After that, you talk to your AI the same way you already do, except now it has access to your real data.
